Before Anyone Can Tell You Yes

Dextr AI

Hotels rarely lose a booking because nobody cared. They lose it when a guest asks a question at 11:47 p.m. and the answer is trapped between a ringing phone, an understaffed desk, a property-management system, and somebody's memory.

Dextr AI has raised a $6.7M Seed round led by Elevation Capital, with Foundation Capital participating. It is the Bay Area company's first institutional financing, and the useful part of the story sits below the chatbot layer.

Co-founders Sajid Shariff, CEO, and Scott Arnold, CTO, are building connected agents for the work that keeps a property moving: reservations, guest requests, housekeeping, staff coordination, group sales, marketing, and back-office operations.

Daisy handles voice reservations. Alfred manages guest communication. Manny works through staff operations. Doss lets employees direct the system by voice or text without turning every new hire into a software archaeologist.

That connection matters. A guest declining housekeeping should change the room schedule. A late arrival should connect reservation data, access instructions, and escalation. A missed call is not just a support metric when the caller was ready to book.

Dextr says its agents handle more than 1M monthly interactions across hundreds of properties. It reports $80K in incremental revenue in 60 days, an 8% EBITDA improvement with 1,200 labor hours removed at a six-property group, and $120K in annual savings while guest satisfaction held at 92%.

Those are company-reported outcomes, not audited benchmarks. They explain why this round attracted attention. Dextr's forward-deployed engineers stay close to operators, configure agents around actual workflows, and remain accountable until the economics show up.

That approach costs more than shipping a login and hoping the customer figures it out. Hospitality may require it because every property brings a different mix of systems, staffing, brand rules, guest expectations, and operational scar tissue.

The $6.7M will fund more forward-deployed engineers, integrations, agents, and deployments. The difficult handoff now belongs to Dextr: turn property-specific implementation into a repeatable platform without sanding away the detail that produces the return.

Hotels have spent years asking people to carry context between software systems. Dextr is financing the bet that the software can finally carry more of that context, while the people get back to the part guests actually remember.

Deeper Analysis 👇 https://devcuration.com/articles/dextr-ai-raises-6-7m-for-hospitality-ai-operations

AMD / World Labs

Silicon roadmaps are asked to predict the workloads that will matter years later. AMD just spent $8.2B to put some of the people inventing those workloads inside the company.

AMD has signed a definitive agreement to acquire World Labs, the spatial-intelligence company founded by Fei-Fei Li, Justin Johnson, Ben Mildenhall, and Christoph Lassner. The all-stock transaction is expected to close by the end of 2026, subject to regulatory approvals and customary conditions.

The price gets attention. The feedback loop is the story.

World Labs builds models that generate, reconstruct, and simulate 3D environments from text, images, and video. Marble turns that work into a product. Atlas extends it into world generation, reconstruction, and robotics simulation before the physical world starts charging tuition.

That work pulls on training, inference, memory, networking, simulation, and edge deployment. A robot cannot blame latency after it bumps into something expensive.

AMD already owns pieces across that stack: Instinct GPUs, EPYC CPUs, Pensando networking, ROCm software, and Helios rack-scale systems. World Labs says the companies began working together last year on model training and inference optimization using AMD GPUs. This acquisition turns that partnership into an organizational choice.

Fei-Fei Li is expected to become AMD's EVP and Chief Scientist after closing, reporting to Lisa Su. Justin Johnson and Ben Mildenhall will continue leading the team with her. Research changes a roadmap only when it reaches the people making product decisions.

World Labs emerged from stealth in 2024 with more than $230M, then announced another $1B in February 2026. It has since opened the World API, introduced Atlas, and acquired robotics-simulation company SceniX.

The transaction still has work attached. It has to close. World Labs has to keep its research talent. Atlas needs broader proof. AMD has to show that proximity changes products, costs, or developer outcomes rather than producing another impressive org chart.

AMD does not want to discover the next compute problem after a rival's software stack has already defined it. It wants researchers close enough to show where the bottlenecks, data types, and system demands are moving while the roadmap can still respond.

World Labs is teaching machines how worlds behave. AMD is betting that the lesson should reach the silicon before the market writes the syllabus.

Deeper Analysis 👇 https://devcuration.com/articles/amd-to-buy-world-labs-for-8-2b-to-shape-ai-silicon

Helix Digital Infrastructure / Samsung

Power, chips, cooling, construction, capital. Every AI campus needs all 5 to arrive on the same schedule.

That schedule is where an infrastructure stack becomes an organizational problem. A delayed grid connection can strand servers. Late cooling can hold up the building. A finished shell without fiber is expensive real estate with excellent ventilation.

Samsung Electronics and 5 affiliates are committing $1B to Helix Digital Infrastructure, the company KKR launched in June to finance, develop, and operate AI infrastructure for hyperscalers.

Samsung Electronics accounts for $500M. Samsung C&T, Samsung SDS, Samsung SDI, Samsung Life Insurance, and Samsung Fire & Marine Insurance are committing the other $500M collectively.

This is additional long-duration capital, not Helix's entire funding base. KKR, Kuwait Investment Authority, NVIDIA, and Vistra committed more than $10B at launch. Samsung joins with an unusually broad set of tools behind the check.

Samsung Electronics supplies semiconductors and data center cooling through FläktGroup. Samsung C&T brings engineering and construction across data centers and power generation. Samsung SDS designs and operates data centers and has moved into GPU-as-a-Service. Samsung SDI works on uninterruptible power systems and battery backup.

Helix is led by former AWS CEO Adam Selipsky, with Waldemar Szlezak serving as CIO while leading digital infrastructure investing at KKR. Their model is built around an expensive truth: hyperscalers need power, data centers, fiber, capital, and accountability to stop behaving like separate projects.

The announcements do not disclose valuation, ownership, investment terms, customers, or completed Helix deployments. Strategic breadth looks great in a partner slide and becomes valuable only when it shortens a real project schedule.

Samsung can bring memory, construction, cooling, data center operations, and backup power into the same development conversation. Helix brings the long-duration capital and coordination layer. NVIDIA and Vistra bring technology and power relationships already embedded in the founding structure.

The opportunity is larger than supplying the AI buildout. It is reducing the number of handoffs capable of delaying it.

Helix now has more capital and another industrial system inside the room. The work ahead is making those capabilities arrive at a hyperscaler's site as one plan, with fewer gaps for time and money to disappear into.

Deeper Analysis 👇 https://devcuration.com/articles/samsung-commits-1b-to-helix-ai-infrastructure

Menos AI

A portfolio manager can explain why a position exists. Six months later, the evidence may be scattered across research notes, market data, risk systems, meeting memory, and the analyst who changed firms.

Menos AI has raised a $5.1M Pre-A round led by returning investor Copper Sky Capital, with Alpha Square Group participating. The financing brings total funding above $10M and backs a question institutional AI must answer: does the system understand the firm's judgment, or merely the documents around it?

Co-founders Junchen (William) Wu, CEO, Baojiang (Chris) Yang, Chief AI Scientist, and Xiang Pan, CTO, are building for that gap. Their Sonαr research agent and AgentOS platform connect internal research, trusted market data, portfolio context, assumptions, and workflows into a private context layer for people and agents.

That sounds technical because the problem is human. Investment reasoning rarely lives in one clean memo. A thesis changes in a risk meeting, absorbs a portfolio constraint, gets challenged by new data, and survives in the memory of people who watched it develop. The firm can keep every file and still lose the logic.

Menos AI wants to make that logic traceable and reusable. Narrative Architecture organizes market themes. Voice Score evaluates which contributors have been early or accurate. Source tracing, data lineage, single-tenant infrastructure, and human review are designed to keep agents inside an institutional process instead of letting fluent output impersonate one.

The company says it works with large asset managers and sophisticated hedge funds. Through Fintech Sandbox, William J. Wu said Menos AI supports teams managing trillions of dollars. Customer names, revenue, retention, and audited performance remain undisclosed, so the signal is adoption interest rather than a victory lap.

Jack Selby led the investment for Copper Sky, doubling down on the belief that models will commoditize while proprietary context becomes more valuable. Alpha Square Group joined a market where the durable asset may be the reasoning a firm can keep, test, and govern.

The capital gives Menos AI more room to build, but the commercial work is specific: prove that institutional memory can survive people, model changes, and portfolio cycles without turning judgment into generic automation.

The next generation of investment infrastructure may be judged less by how confidently an agent answers and more by whether the institution can still explain why it trusted the answer.

Deeper Analysis 👇 https://devcuration.com/articles/menos-ai-raises-5-1m-for-institutional-ai-context

RiskScout

RiskScout's cap table is starting to resemble the market it sells into: venture capital, community-bank funds, and bank innovation networks working on the same financial-crime problem. In regulated software, distribution knowledge belongs beside product knowledge.

The Austin company has closed an undisclosed round led by LiveOak Ventures, with Castle Creek Launchpad, the Bankers Helping Bankers Fund, Alloy Labs, Brue2, and C3 Ventures participating.

RiskScout says revenue more than tripled and the number of banks and credit unions it serves grew by over 250% in the past year. Its customers range from institutions below $100M in assets to organizations above $10B. Those are company-reported figures. The amount, valuation, and formal series remain undisclosed.

Fraud teams look for suspicious behavior and immediate loss. BSA/AML teams carry customer-risk profiles, due diligence, investigations, and filings. Different mandates create different tools and handoffs even when both teams are looking at the same customer.

RiskScout brings those workflows into one platform. Its automation supports alert investigation, customer risk profiling, enhanced due diligence, and SAR and CTR filing preparation while connecting with core banking providers.

Co-Founder and CEO Justin Fischer and Co-Founder and CTO Josh Keys are building with former BSA officers, bank examiners, fraud specialists, and regulators. That experience matters when an investigator or examiner asks why a system made a decision.

The investor list carries distribution logic. Castle Creek Launchpad and the Bankers Helping Bankers Fund were built around technology for community institutions. Alloy Labs works inside the bank innovation ecosystem. LiveOak Ventures leads a group that understands regulated customers cannot buy software like everybody else.

RiskScout already has a live channel test. DCI named it the preferred BSA/AML and fraud provider as DCI prepares to retire BSA Navigator at the end of 2026. That transition turns integration, implementation, support, and trust into more than product-page language.

The new capital will fund team growth, intelligent automation and AI-driven agents, deeper core-provider integrations, and broader bank and credit-union reach.

The work ahead is keeping one risk story intact from alert to investigation to filing while the customer base grows. Every handoff RiskScout removes should give a financial-crime specialist more time to understand the case instead of reconstructing the system.

Deeper Analysis 👇 https://devcuration.com/articles/riskscout-funding-unifies-bsa-aml-and-fraud

Atomic

Inventory becomes expensive twice.

First when a company buys it. Then again when planners have to explain why it is sitting in the wrong place.

Atomic just raised a $12.5M Series A to go after that second cost. The round was co-led by Klass Capital and Madrona, with continued participation from Alumni Ventures and Sandberg Bernthal Venture Partners.

The interesting part is not the size of the round. It is what Atomic wants the product to become.

Most supply-chain tools stop at a forecast, dashboard, or recommendation. A planner still has to turn that output into a purchase order, inventory transfer, production decision, or escalation.

Atomic is building toward a control system that connects business objectives to daily operating decisions. In practical terms, that means the software can handle more routine calls while a person keeps judgment over the decisions that carry real consequences.

That changes the buying question.

Forecast accuracy still matters. But once software starts acting on a recommendation, teams also need to know:

Can the system explain why it made a decision?

What happens when the underlying data is incomplete?

Which exceptions stop the automation and bring in a planner?

Who owns the result when an automated action affects cash, service levels, or a supplier commitment?

Atomic's founders learned this problem in a demanding environment. Michael Rossiter, Neal Suidan, and Jeff Goodrich worked together around Tesla's Model 3 launch, where planning had to absorb rapid production changes and an unusually high cost of being wrong.

The company now names DoorDash DashMart, HelloFresh Good Chop, Starface, OOFOS, LMNT, and Vincero as customers. It says onboarding can happen in as little as 30 days, and its published customer stories point to lower inventory, higher in-stock rates, and less planning work. Those are company-curated customer results, not universal benchmarks, but they show the operating outcome Atomic is chasing.

This round gives the team room to move from decision support toward execution. It also raises the evidence bar.

Adrian Schauer of Klass Capital and Matt McIlwain of Madrona are joining Atomic's board alongside Jon McNeill of DVx.

The next proof point is not another AI feature. It is a growing set of routine decisions that customers are willing to delegate because the data, guardrails, and exception paths are good enough to trust.

That is where supply-chain software becomes operating infrastructure.

Deeper Analysis 👇 https://devcuration.com/articles/atomic-raises-12-5m-in-series-a-funding

EliseAI

The hardest permission in enterprise software is not access to the data. It is permission to act inside the workflow that keeps an apartment building or medical practice moving.

EliseAI has raised $350M at a $4B valuation. Andreessen Horowitz, Bessemer Venture Partners, and Ontario Teachers' Pension Plan led the financing, with Sapphire Ventures and Navitas Capital participating.

That investor group is underwriting a company that has spent 9 years turning small permissions into operational authority. Customers first adopted EliseAI to handle leasing and resident communications. The platform now reaches maintenance, renewals, collections, tours, scheduling, intake, and prior authorizations.

Minna Song, co-founder and CEO, and Stoyan (Tony) Stoyanov, co-founder and CTO, built EliseAI around a practical reading of vertical AI: fluency gets attention, but reliability earns more of the job. The company says it now powers one in five U.S. apartments, has surpassed $200M in ARR, and has doubled revenue year over year for five consecutive years. Those are company-reported metrics, but they describe the operating base behind the valuation better than another clever AI demo could.

Apollo makes the direction especially clear. Instead of sitting beside the workflow as a chatbot, the agent works inside EliseCRM and related applications, inherits a user's permissions, and can act across messages, settings, tasks, and dashboards. EliseAI says the system has been tested across more than 7,000 evaluations.

Housing gave EliseAI years of conversations, edge cases, integrations, and operator trust. Healthcare raises the consequence. Scheduling, intake, prior authorizations, billing, and patient questions carry different rules, but the commercial challenge is familiar: the software has to understand the work well enough to help without making the next human repair its mistakes.

The company plans to use the new capital to expand the workflows its AI can automate, accelerate product development, and grow engineering, deployment, and sales. That matters because vertical AI does not become infrastructure when it gives the best answer. It becomes infrastructure after customers keep granting it the next permission.

At $4B, EliseAI has enough capital and customer reach to ask for more of those permissions. The useful story will be written inside the buildings and practices where the system has to carry context, policy, and judgment from one action into the next.

Deeper Analysis 👇 https://devcuration.com/articles/eliseai-raises-350m-at-a-4b-valuation

Elio Mortgage

Oren Michaely and Arad Lev Ari made the expensive choice early: if they wanted to rebuild mortgage origination, they had to own the mortgage company too.

That decision separates Elio Mortgage from tools asking lenders to add another dashboard to a process already held together by handoffs. A borrower sees one home loan. The operation sees applications, documents, pricing, disclosures, conditions, title, insurance, underwriting, deadlines, and a small village waiting for the same file to tell the truth.

Elio Mortgage emerged from stealth with $5.1M in Pre-Seed funding led by Motive Partners and Social Leverage. Jeff Horing, co-founder and managing director of Insight Partners, also participated as an angel investor.

Oren Michaely brings applied AI experience from Microsoft and Motive Partners. Arad Lev Ari brings real estate investing and finance experience from KKR and Deutsche Bank. Chief Brokerage Officer Steven Carey brings the less glamorous credential: he has owned a brokerage and produced loans.

That mix matters because mortgage software rarely fails from a shortage of screens. It fails in the space between them, where somebody has to notice the missing document, interpret the exception, chase the condition, explain the delay, and keep the borrower from feeling like the largest financial commitment of their life has disappeared into an inbox.

Elio is building one shared loan record with AI agents working across onboarding, processing, closing, and post-close service. Agents prepare work, coordinate follow-ups, and flag risk. Loan officers keep the judgment, advice, and relationship.

Elio is also turning that stack outward through Elio Embedded. Financial advisors, real estate agents, home builders, and institutional owners can offer mortgage services without assembling the licensed operation themselves. Distribution arrives with trust already attached, while Elio carries the machinery.

The round will fund product development, licensed expansion, and loan-officer recruitment. Elio's site lists 37 loan officers across 22 states. Those figures and the platform's efficiency claims are company-reported. No loan-volume, cost-savings, close-time, revenue, or valuation metric was disclosed.

Elio now has to prove that owning the workflow lets its technology remove coordination cost without sanding away the human judgment a regulated mortgage still requires. If the work can recede while the relationship stays visible, the economics may finally change where the borrower can feel it.

Deeper Analysis 👇 https://devcuration.com/articles/elio-mortgage-raises-5-1m-to-rebuild-origination

Reco

A quarterly access review can be accurate on Monday and obsolete by Friday once AI agents start chaining applications together.

The app was approved. The employee was approved. The service account was approved. Then an agent inherited the permissions, connected another tool, opened a browser session, and turned 4 acceptable decisions into 1 risk nobody had reviewed as a whole.

That moving chain is the problem Reco is financing itself to solve.

Reco has raised $55M in additional funding, bringing its disclosed total to $140M. The financing includes a strategic investment from AT&T Ventures, with support from Forestay and Quadrille Capital. No round label or valuation was disclosed.

AT&T matters here as more than a logo on the cap table. Reco says the telecom company is both a customer and investor. That puts the wager inside a real enterprise environment where agents touch applications, identities, data, permissions, and workflows that already carry consequences.

Ofer Klein, CEO, Gal Nakash, CPO, and Tal Shapira, CTO, founded Reco in 2020. Their platform uses the Reco Graph to map what agents can reach, which identities they rely on, what tools they connect, and how access can spread across an operating ecosystem.

Reco says the platform now supports more than 280 app integrations and 1,000 detection controls. Those are company-reported figures, but the architecture points at the right problem. An agent can look low risk in isolation while its OAuth grants, service accounts, browser sessions, MCP tools, and connected applications create a very different blast radius together.

The company raised a $30M Series B in February, taking total funding to $85M. Seven months later, this $55M addition is aimed at sales, partnerships, channels, and customer support. That use of funds says the next constraint is not only building more controls. It is helping large enterprises deploy them without turning every AI project into a security veto.

That is a different scale-up problem. More integrations create coverage. More customers create implementation work, support demand, partner coordination, and an obligation to explain risk without flooding teams with another alert queue.

Security teams have spent years learning how to govern employees, endpoints, cloud workloads, and SaaS applications. Agents borrow pieces from every one of those systems, then move across the seams where ownership gets fuzzy.

Reco now has the capital to make those seams visible while customers are still deciding who owns them.

Deeper Analysis 👇 https://devcuration.com/articles/reco-raises-55m-to-scale-agent-security

MoneyStack

A counseling appointment can help someone stop gambling. The credit-card balance, overdue taxes, household tension, and urge to chase the loss are still waiting when the session ends.

Alex De Marco built MoneyStack around that unfinished part of recovery. The company puts financial counselors inside behavioral-health programs so money recovery stays with the care plan.

MoneyStack has raised $1M in new funding led by Cistern Capital. The American Heart Association's Social Impact Funds, O'Shaughnessy Ventures, Sidecut Ventures, Gaingels, and angel investors also participated. FinSMEs reports that the round brings total funding to nearly $2M. The stage and valuation were not disclosed.

The service starts with the work people need when gambling has damaged the household balance sheet: virtual counseling, debt and spending plans, group sessions, and a personalized Financial Recovery Plan. Through GamFin, MoneyStack says state partners in 19 states fund counseling.

That payer structure matters. A state agency or behavioral-health partner funds the service because financial distress can keep pressing on treatment, family stability, and the temptation to chase losses. MoneyStack sells infrastructure to a care system while delivering a human intervention to the person living with the damage.

The new capital will support more technology. FinSMEs says MoneyStack plans to turn years of counseling experience and client data into tools that can identify financial harm earlier. GamFin Guardian, currently on a waitlist, shows the direction: connect financial accounts, flag gambling-related activity, and notify trusted people selected by the user.

Transaction data can reveal trouble before somebody is ready to describe it. It is also deeply intimate. The software has to be accurate, consent-driven, secure, and calm enough to help without making a person in recovery feel watched by another system.

Cistern Capital's wellness thesis spans financial, emotional, physical, and social health. The American Heart Association's Social Impact Funds backs health-care access and economic vitality. MoneyStack sits where those mandates stop being separate investment categories and become the same client problem.

The round gives Alex De Marco's team more room to build between the counseling session and the next financial decision. MoneyStack now has to preserve that room's trust while helping more programs, providers, families, and clients carry the plan into the week.

Deeper Analysis 👇 https://devcuration.com/articles/moneystack-raises-1m-for-financial-recovery

RemoteThreat

A machine can execute an offensive cyber task. It cannot accept the consequence when that task crosses the line.

That distinction sits at the center of what Chris Thompson and Shawn Jones are building at RemoteThreat. The company has disclosed $7M in equity financing while bringing an integrated Offensive Cyber Operations platform into the market for advanced red teams, U.S. government mission teams, and vetted defense partners.

The SEC filing says RemoteThreat sold $7M of a planned $7.3M offering to 2 investors, with the first sale on January 16, 2026. It names no lead, syndicate, valuation, or round stage. DataTribe publicly calls RemoteThreat a portfolio company; the other investor is undisclosed.

The money is only useful if the operating idea survives contact with the work.

Chris Thompson spent years leading IBM X-Force red-team and adversary-services operations. Shawn Jones led capability R&D and worked as a red-team operator inside IBM X-Force. They are turning that experience into the O/C/O Platform, which connects mission planning, initial access, implants, command and control, obfuscation, analysis, and bounded AI workflows.

Offensive operations rarely fail from a shortage of tools. They fail in the handoffs between objective, access, tasking, evidence, approval, and the operator who must explain what happened.

RemoteThreat is financing the argument that those handoffs belong in one operational picture. Its platform is designed to keep rules of engagement, approvals, auditability, and No-Strike controls connected across each consequential handoff while allowing human operators to delegate defined work to machines.

That is more serious than asking whether a model can write an exploit. Speed without scope is a liability. Autonomy without someone who can stop the operation is somebody else's incident report waiting to be written.

RemoteThreat still has plenty to prove. It has not publicly disclosed revenue, customer counts, independent performance results, or detailed use of proceeds. Its product claims await market validation.

Still, the architecture reveals the bet. AI may accelerate offensive tradecraft, but responsibility cannot be outsourced with the task. Chris Thompson and Shawn Jones are building for the moment when red teams need machines inside the operation without letting them own it.

The software can move first. The operator still has to know why.

Deeper Analysis 👇 https://devcuration.com/articles/remotethreat-discloses-7m-for-offensive-cyber-operations

ai3Bio

Autoimmune care can ask a patient to manage the same underlying biology for years, one suppressed signal at a time.

ai3Bio is financing a different bet: identify the immune cells driving the damage, then remove those cells directly while preserving the rest of the immune system.

The Watertown, Massachusetts biotech launched with a $48M Series A led by founding investors UPMC Enterprises and Ziff Capital Partners, with Cockrell Interests and Tanis Ventures participating. CEO Steven Altschuler is leading the company alongside scientific co-founder Jonathan Kagan, CTO Michael Lyman, and Chief Financial and Business Officer Nick Seaver.

The target is Th17 cells, a subset of T cells implicated in autoimmune inflammation. Many widely used therapies interrupt cytokines such as IL-17, IL-23, or TNF after those signals are produced. ai3Bio is developing 2 preclinical platforms intended to remove pathogenic Th17 and other CD161-positive T cells at their source.

STARx packages engineered mRNA inside an antibody-guided lipid nanoparticle. Once delivered to the targeted T cell, the payload is designed to activate the cGAS-STING pathway and trigger apoptosis, the cell's programmed cleanup process. T Deplete takes another route, using an antibody to mark the cell population for elimination through antibody-dependent cellular cytotoxicity.

That dual-platform design matters because biology rarely rewards a single delivery mechanism with universal permission. ai3Bio can pursue the same cellular thesis through 2 therapeutic architectures while learning which diseases, patients, and safety constraints fit each one.

The company says patient-blood studies showed its LNP technology removed disease-causing Th17 cells, while its antibody platform produced efficacy in a non-human-primate autoimmune model. Those are preclinical results, not evidence of benefit in people. IND-enabling work is underway, and ai3Bio is targeting an initial patient trial in 2027.

UPMC Enterprises brings a health-system commercialization engine. Ziff Capital Partners brings company-building capital and leadership. The round is paying for the difficult handoff from an elegant immunology thesis to a therapy that can be manufactured, dosed, monitored, and trusted in humans.

For the more than 15M Americans estimated to live with at least 1 autoimmune disease, durable remission is the ambition. ai3Bio now has to prove that selective T-cell depletion can make that ambition clinically precise without taking healthy immune function with it.

Deeper Analysis 👇 https://devcuration.com/articles/ai3bio-raises-48m-to-target-autoimmune-t-cells

Tiny Health

Microbiome testing has become very good at producing detail. The business gets harder when a customer, clinician, or researcher asks which part of that detail deserves action.

Tiny Health has closed a $33M Series B led by B Capital, bringing its company-stated total funding to $46M. The syndicate includes 6 new firms and returning backers Spero Ventures, TheVentureCity, and Overwater Ventures.

Cheryl Sew Hoy started Tiny Health while searching for better answers about her daughter's eczema and food sensitivities. It launched for mothers and babies, then followed the demand outward.

Tiny Health now sells gut and vaginal microbiome testing across life stages. Powered by Tiny supplies testing, analytics, and research infrastructure. The company says it has served nearly 200,000 customers and 6,000 practitioners.

Beauty Independent reports that annual revenue has crossed eight figures and is at least doubling, with roughly half from the business arm. Adult testing has passed pediatric testing by volume. Those company-reported metrics show how far Tiny Health has moved beyond its original wedge.

The round is financing that broader platform. Tiny Health is introducing TinyAI, trained on what it describes as nearly 200,000 proprietary profiles, plus Gut Microbiome Age and Gut Resilience metrics. It is also committing $5M to a research program.

That commitment belongs near the center of the story. A 2025 company-sponsored randomized controlled trial of 54 C-section-born infants reported lower odds of atopic dermatitis in the intervention group. The study had a small sample, short follow-up, parent-reported outcomes, and extensive Tiny Health affiliations among the authors. Promising evidence still carries limits.

Tiny Health says its tests are not diagnostic and do not replace medical care. CMO Elisa Song has argued that the field needs more evidence before microbiome testing becomes standard care. That restraint matters when a report can look more settled than the science behind the next decision.

B Capital is backing consumer traction, practitioner distribution, enterprise relationships, and a growing dataset. Tiny Health now has to make those assets reinforce one another: more use should produce stronger studies, clearer interpretation, and fewer reasons for clinicians to treat microbiome results as expensive curiosity.

Its next chapter will be decided in the distance between a sophisticated report and a result a clinician can defend when a family asks what to do on Monday.

Deeper Analysis 👇 https://devcuration.com/articles/tiny-health-raises-33m-for-microbiome-intelligence

Destro AI

Warehouse robots are getting better at their own jobs. The warehouse still needs somebody to decide which job comes next.

That decision has to account for orders, inventory, dock conditions, labor availability, equipment state, and the thousand small exceptions that never appear in a clean product demo. Add autonomous mobile robots, forklifts, arms, and people from different vendors, and the hardware can be capable while the operation remains fragmented.

Destro AI has raised an $8M Seed round to build the intelligence layer above that fragmentation. Base10 Partners and Bonfire Ventures co-led the financing, with CoFound Partners participating.

Founder and CEO Manthan Pawar and co-founder and CTO Soorya Narayanan are building Destro around a blunt idea: warehouse operators do not need every machine to come from the same company. They need the machines, people, and enterprise systems already on the floor to work from the same operating picture.

MothershipOS handles intelligence outside the robot. It monitors conditions, weighs priorities, and coordinates tasks across robots and associates. VisionOS handles intelligence on the robot, using models trained from human demonstrations to support perception and manipulation. One manages the operation. The other helps the machine deal with the physical world.

The $8M will support enterprise deployments, product development, and growth across engineering and research. Destro says it is already generating commercial revenue and running in production with large logistics operators.

Yusen Logistics is deploying Destro in transload operations, starting with cart movement as freight volume, labor, staging, and dock conditions change. Yusen Director of Automation Rick Brunelle says the platform is giving teams more intelligence and control.

That is the line between an impressive robot and useful warehouse infrastructure. The demo proves a machine can perform a task. Production asks whether the entire floor can absorb exceptions, preserve service levels, and keep moving when the plan gets old five minutes after the shift starts.

Base10's Adeyemi Ajao and Bonfire's Jennifer Richard are backing the deployment problem. Destro now has to turn mixed fleets and messy operations into a system customers can trust across vendors and sites.

The capital buys more deployments. The warehouse floor will decide whether one shared mind can keep earning responsibility for what happens next.

Deeper Analysis 👇 https://devcuration.com/articles/destro-ai-raises-8m-for-shared-warehouse-intelligence

Kahua

Scott Unger and Brian Moore had watched a construction-software company reach Autodesk. When they started again, they chose to build a system that could bend around how owners govern capital programs without letting every project invent reality.

That decision is now carrying a valuation above $1B. Kahua announced a minority growth investment from Bain Capital Tech Opportunities after reaching $100M in company-reported annualized revenue. The investment amount, ownership, board rights, and detailed terms were not disclosed.

Construction software sounds simple until concrete, contracts, public money, contractors, and a 10-year asset lifecycle enter the same meeting. Teams need different approval paths, while leadership still needs one defensible record when the schedule slips or the budget moves.

Scott Unger, co-founder and CEO, and Brian Moore, co-founder and president, designed Kahua so organizations can configure apps, workflows, permissions, funding controls, reporting, and handover while keeping the data connected.

Kahua now reports 2,500+ customers, 500+ certified partners, 70K+ projects, and more than $400B in capital programs managed on the platform. Those are company figures, but $100M in annualized revenue gives the configurability thesis a commercial receipt.

The next layer is AI, where construction technology stops being a demo business. Noa works inside Kahua's governed environment. kBuilder Canvas lets teams design workflows with natural language or a visual builder. kCapture connects field imagery to the project record. CTO Colin Whitlatch and the product team are putting intelligence inside permissions, history, and context.

That distinction matters in federal, transportation, healthcare, education, energy, and data-center programs. A generated answer can be useful. A generated action has to know who is allowed to make it, which record controls, and what an auditor will ask 3 years later.

Bain is backing the next phase across AI, product innovation, go-to-market, customer success, and talent. The valuation recognizes the scale already built. The investment raises the standard for turning governed data into faster decisions without turning flexibility into another form of sprawl.

Kahua's first act helped construction move project controls online. Its second has spent 17 years making enterprise software conform without losing the record. The work ahead is to make intelligence move through that foundation with the discipline its customers' capital requires.

Deeper Analysis 👇 https://devcuration.com/articles/kahua-bain-growth-investment-1b-valuation

Dodge AI

Enterprise software rarely breaks because nobody knows how to fix it. It breaks slowly because the person who remembers why the exception exists is on another team, at another vendor, or halfway out the door.

That memory becomes a ticket and then a consultant call. The fix works, but the reason disappears into a queue, a configuration note, or somebody's head. Six months later, the company pays to rediscover its own system.

Rebhav Bharadwaj, CEO, and Aditya Thakur, CTO, built Dodge AI around that leak. The company has raised a $2.65M Seed backed by Accel and Google's AI Futures Fund, with Schema Ventures, New Build Ventures, Antler, and angels from the SAP ecosystem participating.

Dodge AI is building an AI control plane for enterprise application maintenance. Its agents sit across SAP, Salesforce, Microsoft Dynamics, Kinaxis, Oracle JDE, ticket history, and business processes to trace failures, recommend fixes, and preserve the exception logic for the next incident.

The important word is context. An ERP estate is a living archive of choices the standard product never anticipated. Pricing exceptions, warehouse rules, approval paths, and local workarounds accumulate until the real operating manual is spread across consultants and memory. A fluent model can still be dangerously wrong if it cannot reconstruct that environment.

Dodge AI says it works with more than a dozen enterprises, half publicly listed, and handles hundreds of queries every hour. Its CPG case study reports 30% fewer MRP support tickets and 11 planner hours saved per plant each week. The results are company-reported and the customer is unnamed, but they reveal the wedge: start with the incident queue, where the pain is measurable and cannot wait for a transformation program.

That also explains the investor logic. Accel and Google are not financing a prettier chatbot for SAP. They are backing the possibility that maintenance knowledge can become software infrastructure, then compound into the context agents need before enterprises will let them act in production.

Dodge AI has to earn permission inside systems that move inventory, money, orders, and customer commitments. Every resolution adds context. Every wrong action becomes an operating event somebody has to explain.

If Dodge AI can make those decisions traceable, maintenance becomes a system that remembers why the enterprise works the way it does. That memory is where the next ticket gets cheaper, the next change gets safer, and the consulting handoff starts to move.

Deeper Analysis 👇 https://devcuration.com/articles/dodge-ai-raises-2-65m-for-erp-maintenance-agents

ITpipes

A city council cannot approve a rehabilitation budget from a folder of pipe videos.

Every crack, root intrusion, offset, and failed joint has to become standardized condition data. That record must survive a trip from a CCTV truck to an engineer, a GIS map, an asset-management system, a capital plan, and the public meeting where someone asks why this pipe gets fixed first.

ITpipes has added Luminate Capital Partners to that chain. The Seattle infrastructure-software company announced a strategic growth investment to accelerate product and AI development for North American wastewater and stormwater operators. The amount, valuation, and terms were not disclosed.

Founders Cori Criss and Al Rossmeisl built around a stubborn reality: inspection data becomes valuable when it fits how crews, engineers, and agencies work. CEO Ned Bebawy and CTO Anshuman Srivastava are carrying that logic into AiDetect, pairing machine analysis with review by NASSCO-certified experts instead of treating an algorithm like a clairvoyant intern.

ITpipes says AiDetect reaches 97% defect-detection accuracy while cutting coding and QA time by at least 50%. More than 300 utilities, municipalities, and CCTV contractors use the software. Those company-reported figures frame Luminate partner Andrew Baldwin's wager: underground infrastructure needs a trustworthy operating record, not another clever demo.

FieldVision captures field work. CoreVision organizes inspection data, maps, media, and reports in the office. SmartVision connects the 2, while Esri ArcGIS and asset-management integrations carry the record into planning. The capital is meant to bring inspections, coding, and asset history together for prioritization and capital decisions.

The timing is not subtle. The EPA reports at least $630.1B of U.S. clean-water infrastructure needs over 20 years. ASCE gives wastewater infrastructure a D+ and estimates a $69B annual wastewater and stormwater capital gap. Agencies must inspect more pipe and defend more spending without discovering a spare workforce behind the treatment plant.

ITpipes raised $20M from Trilogy Search Partners and Miramar Equity Partners in 2024. Luminate's undisclosed investment adds a growth-software specialist and raises the obligation. Faster coding must produce records that stay consistent across crews, contractors, years, and systems.

The camera shows what is underground. ITpipes is building the chain of custody that decides what a community does about it.

Deeper Analysis 👇 https://devcuration.com/articles/itpipes-luminate-growth-investment-underground-data

Optikos / Omega Optical

Optical failures have a favorite hiding place: the space between suppliers.

The design house meets its model. The coating vendor hits the transmission curve. The manufacturer ships an assembly on tolerance. Then the finished system misses performance, and everybody arrives at the meeting with a passing report.

Optikos has acquired Omega Optical to put more of that argument under one roof.

The September 29 transaction combines Optikos's engineering, manufacturing, and metrology with Omega's coatings, filters, gratings, infrared optics, and assemblies. Both were already owned by Artemis, making this a bet on one accountable path from optical concept to verified system performance.

That matters in aerospace, defense, life sciences, and semiconductor manufacturing, where programs can cross design, fabrication, coating, assembly, and test before the customer knows whether the system works. Each handoff adds schedule and another place for responsibility to get refracted.

David Cooper, previously Omega Optical's CEO, now leads the combined Optikos. Rudi Coetzee becomes executive chair. Stephen D. Fantone remains founder, strategic adviser, and chair emeritus, while CTO David Imrie carries Optikos's technical continuity.

The deal terms were not disclosed. No purchase price, consideration structure, regulatory process, or separate close date was published. The companies did say the acquisition is complete and the leadership change is effective immediately.

Customers were given a practical promise: existing programs, contracts, schedules, engineering contacts, and product roadmaps will continue through the transition. Artemis is committing growth capital to manufacturing capacity and accelerating hiring.

That investment is where the acquisition either becomes useful or stays a capabilities slide.

Optikos was founded in 1982 and grew from engineering into manufacturing and metrology. Omega began with filters in 1969, then expanded through coatings, gratings, infrared components, and assemblies. Their operating thesis now meets at one point: customers buy a system that has to perform after every specialty has touched it.

The combined company now owns more of those handoffs. It also owns more of the explanation when a coating, component, assembly, and measurement do not agree. The first shared programs will show whether one owner can turn that accountability into faster decisions without sanding away the depth that made both companies worth combining.

Deeper Analysis 👇 https://devcuration.com/articles/optikos-acquires-omega-optical-workflow

flowscope

Every company has a process that works because one person remembers which invoice gets an email, which exception needs a spreadsheet, and which approval needs a call.

That knowledge rarely survives in the process map. It lives in the workaround nobody documented and the handoff that only makes sense after an operator explains it.

flowscope has raised a $3.8M Seed to turn that hidden operating layer into working AI systems. Y Combinator, Red Forest Capital, Anderson Growth Partners, and Outbound Capital backed it, with Latitude Fund, X26, and angels participating. No lead or valuation was disclosed.

CEO Samuel Mirpuri saw the delivery problem while leading digital-transformation programs at McKinsey. CTO Javier Leguina brings engineering experience from ModelML, Encord, and UCL. Their company lives inside the distance between a transformation plan and the system that has to run on Monday.

flowscope’s agents observe how work moves, map the process, redesign the flow, and deploy automations over existing tools. The pitch is aimed at CFOs, COOs, and private-equity teams with no shortage of AI demos and no appetite for another deck.

The company says it is embedded in more than 10 businesses and works with major US private-equity firms. The customers are unnamed and the outcomes are company-reported, but buyers are letting flowscope watch work and put software into production.

That permission matters. McKinsey found 88% of surveyed organizations used AI in at least 1 function in 2025, yet nearly two-thirds had not begun scaling it across the enterprise. Model access spread quickly. Operational context and accountability did not arrive in the same download.

The investor group is backing that implementation layer. Red Forest, Anderson Growth Partners, Outbound, and Latitude bring different family-office and early B2B perspectives. Y Combinator placed flowscope in its Spring 2026 cohort.

The $3.8M gives flowscope room to hire in San Francisco and deepen a delivery loop that still has to prove it can compound. Each deployment creates knowledge about exceptions, permissions, and write-back patterns. Value builds when that knowledge improves the next implementation without turning customer context into a template.

Samuel Mirpuri and Javier Leguina now have to make observation repeatable without making it shallow. The product earns its place when hidden process knowledge can move safely through software while the operator keeps the judgment the system has not earned.

Deeper Analysis 👇 https://devcuration.com/articles/flowscope-raises-3-8m-to-automate-enterprise-work

Efficient Computer

The chip industry's favorite energy fix is specialization: make one workload fast, give it a dedicated accelerator, and let the rest of the application keep paying the power bill.

Efficient Computer is financing a broader answer. The Pittsburgh company announced agreements for more than $97M in Series B financing at a $650M valuation, led by TQ Ventures. Eclipse, Union Square Ventures, Giant Ventures, Triatomic Capital, TO Capital, TF Capital, Mana Ventures, Toyota Ventures, Overmatch, and Borderless also participated.

That syndicate is financing 2 jobs that rarely arrive in the same round. Efficient plans to ship its Electron E1 processor in volume while scaling the same Fabric architecture toward datacenter-class performance.

Brandon Lucia, CEO, Graham Gobieski, CTO, Nathan Beckmann, Chief Architect, and co-founder Alex Hawkinson built Efficient around a stubborn choice: preserve general-purpose programmability while cutting the energy spent moving data. The effcc Compiler accepts familiar C and C++ workflows, then maps them onto a spatial dataflow architecture.

Physical AI is not one neat model running alone. A robot, infrastructure sensor, spacecraft, or wearable may need perception, signal processing, control, communication, and ordinary application logic inside one power budget. Optimizing only the AI kernel leaves too much of the system untouched.

The round creates a larger obligation. Efficient reports 10-100x energy improvements and targets more than 10x lower energy consumption as Fabric moves toward datacenter-class systems. Those are company claims. The next proof has to come from repeatable deployments across a wider performance range.

An August SEC filing showed $93.1M sold toward an open $103.1M offering. The September announcement supplies the commercial record: Series B, TQ Ventures as lead, a $650M valuation, participating firms, and agreements exceeding $97M. Efficient says total funding now stands at $173M after a $16M Seed and $60M Series A.

Lawrence Berkeley National Laboratory estimates data centers could consume 11.8% of U.S. electricity by 2030. On a robot or wearable, the same constraint shows up as heat, battery life, or operating time.

Efficient is trying to make one architecture credible at both ends of that power problem. This capital buys the chance to prove that whole-application efficiency, familiar software, and dependable chip supply can become one purchasing decision from the robot to the datacenter.

Deeper Analysis 👇 https://devcuration.com/articles/efficient-computer-announces-97m-series-b-ai-chips

Enable Ventures

Accessibility has spent decades showing up after the product meeting, usually as a compliance ticket with a deadline.

Enable Ventures just closed $50.31M on the argument that it belongs in the investment memo before the first line of product code.

The vehicle is Enable Ventures Fund I, led by founder and managing partner Regina "Gina" Kline and built within the Sorenson Impact Platform of Funds. The final close brings the Ford Foundation, UnitedHealth Group, Liberty Mutual Investments, Ally Bank, SHRM, Ascension, and other LPs into the same thesis.

That thesis is more demanding than calling accessibility a large market. Enable invests across worktech, digital health, next-generation assistive technology, and founders with disabilities because people who have been excluded from products and capital often see requirements the market has trained itself to ignore.

The fund has made 11 investments. Be My Eyes uses AI for visual interpretation. Kalogon builds sensor-enabled seating. Cionic is developing a wearable neural sleeve. DeepWalk uses mobile LiDAR for sidewalk access. Inclusively works on workplace accommodations.

Different products, same capital question: what becomes visible when disability experience enters the design process early enough to shape the product instead of auditing it later?

Kline came to that question through civil-rights law. Before venture capital, she spent years working on disability employment and helped challenge systems that paid workers with disabilities pennies per hour. Moving upstream into investment changes the available tool. The argument can now be tested through company selection, product decisions, customer adoption, and returns.

The accounting deserves precision. Enable’s 2023 SEC filing named a $75M target. That was not money raised. The final close is $50.31M in commitments, with the announcement body rounding the figure to $50M. Fund economics, check sizes, individual LP commitments, and performance remain undisclosed.

What is disclosed is a portfolio still in motion. Axios reports Enable expects roughly 25 investments from Fund I, leaving the final close closer to a starting gun than a victory lap.

Institutional capital has agreed that disability inclusion belongs in underwriting. Enable’s companies now have to show how that perspective travels from a design advantage into products people choose, organizations buy, and markets reward.

Deeper Analysis 👇 https://devcuration.com/articles/enable-ventures-closes-50-31m-disability-tech-fund