The People Still in the Room

Flexcompute

Zongfu Yu could build a device faster than he could find out whether it worked. Researching silicon optical isolators, he waited days for simulations and spent hours on the physical experiments.

That imbalance helped produce Flexcompute. CEO and cofounder Zongfu Yu built the company with Qiqi Wang and Shanhui Fan; president and cofounder Vera Yang is carrying that ambition into industry.

On October 12, the Boston-area company announced strategic investment from Samsung Ventures and Applied Ventures, the venture arm of Applied Materials, among other investors. The amount was undisclosed. The named backers bring manufacturing experience beside a company giving designers more room to think.

A chip designer can have an interesting idea and spend the afternoon waiting for the machinery that evaluates it. When each calculation carries that much delay, teams have to ration their curiosity. Faster simulation changes the number of alternatives worth exploring before someone commits to a design.

Flexcompute already sells Flow360 for computational fluid dynamics and Tidy3D for electromagnetic simulation. Its newer Physics World Model ambition combines neural networks with first-principles physics. The proposed working arrangement is unusually clear: engineers choose the objective and approve the result; agents handle the iterations.

The manufacturing connection is taking shape in the tools. In March, Flexcompute announced an integration of GlobalFoundries' silicon photonics process stack into PhotonForge. Designers can work with the foundry's layer definitions and material parameters directly, keeping a simulated device connected to the process intended to manufacture it.

In June, the company described agents proposing photonic designs, running simulations and checking fabrication constraints in a loop. These are company-reported demonstrations. The commercial proposition: an engineering team could spend more of its day examining viable options, with the manufacturing rules already inside the conversation.

The new investment is intended to expand research, GPU capacity and coverage across interacting kinds of physics. Its $22M Series B, led by Coatue, dates to 2021; this week's strategic backing arrives as the job of its software expands from answering engineering questions toward participating in the design process.

For Vera Yang and Zongfu Yu, that brings the old waiting problem into a new room. An engineer reviewing an agent's next proposal will need the geometry, the manufacturing assumptions and the calculation to travel together all the way to approval.

Scarlet

James Dewar and Jamie Cox went from building medical AI at Babylon Health to deciding whether other teams' medical AI could reach the market. That is an unusually consequential career move: becoming the institution your former self needed to work with.

At Scarlet, James Dewar is CEO and Jamie Cox is CTO. Their London-based certification body has raised a $50M Series B led by Headline, with new investor Notion Capital and returning backers Atomico, Creandum, Kindred Capital and Entrepreneurs First. The round was announced on October 12.

Medical software brings a recurring obligation into the approval process. A manufacturer improves a model, adds a device or changes a release, and the evidence around that product has to remain useful. The people assessing it need to understand both what changed and what the change means for its regulated use.

Scarlet has built its service around that ongoing relationship. Its public procedures cover technical documentation, audits and surveillance. Headline describes qualified reviewers making regulatory judgments while technology processes and organizes the documentation, with findings tied to evidence and the relevant requirements.

That division of work gives the customer something valuable to plan around. A development team can build an update quickly; the commercial team still needs a credible path to offering it. Review capacity ends up inside the launch calendar, whether anyone put it in the original product roadmap or not.

The authority behind that service took work to obtain. The MHRA designated Scarlet NB UK as an Approved Body focused on software and AI medical devices. The FDA lists Scarlet NB B.V. as a recognized third-party reviewer for eligible 510(k) submissions. Each jurisdiction and certification scope carries its own responsibilities.

Scarlet says customers have cut a year or more from certification timelines and moved update cycles from months to weeks. Those are company-reported outcomes. Its customer stories span AI physiotherapy, radiology and dermatology, giving the service a concrete set of uses beyond a promising slide about healthcare AI.

The funding will support more expert capacity, wider certification offerings and expansion in the US and Asia-Pacific. As Scarlet takes on more manufacturers, James Dewar and Jamie Cox will be carrying more of their customers' release calendars into the same institution they built to keep medical software moving.

Deeper Analysis 👇 https://devcuration.com/articles/scarlet-raises-50m-series-b-for-ai-medical-certification

Switched Source

40 years is a long time to live with a purchasing decision. Charles Murray sees utilities making commitments on that scale while the demand arriving on their circuits keeps changing.

The Switched Source cofounder and CEO is selling those planners another option. Alongside cofounder and COO Lane Nelson, he is building equipment that helps a utility find usable capacity in the wires it already owns, with more than $17M announced on October 13 to expand manufacturing, testing and deployment.

Cubit Capital led the financing, joined by ABB, Emerald Technology Ventures, Red Cedar Ventures, Southern States, Southwire Company and other strategic investors. The announcement leaves the round series and valuation undisclosed. The work includes more production capacity, more utility installations and manufacturing, quality and supply-chain hiring in Chicago.

Phase-EQ shifts power between the 3 phases of a distribution circuit. A heavily loaded phase can constrain the circuit while another has room. Switched Source uses power electronics to rebalance that load dynamically, giving planners a way to recover capacity as charging habits, heat pumps and neighborhood demand change.

The company reports more than 150,000 operating hours across 12 utility service areas, with field data showing 10%-25% gains in load-serving capacity. It puts the typical journey from purchase order to activation at 14-20 weeks. Those are company-reported results and timelines; the feeder's conditions still determine the opportunity.

EPRI and Ameren studied that dependence in an earlier modeling project. Location, load and time of day affected the imbalance. A utility engineer has to decide where to place the equipment and which constraint it can relieve.

Murray and Nelson started Switched Source in 2016 after meeting in a University of Chicago Booth classroom. The company expanded to a 20,000-square-foot Chicago production and testing facility in 2025. The round follows years of utility deployment.

That world changes shape during a storm. Switched Source says a separate 5-year, $99M GRIP project will deploy Phase-EQ for Southeast restoration, helping rebalance circuits as feeders are reconfigured. The project value stays separate from the new financing.

For Murray and Nelson, more factory capacity means more devices arriving at utility sites where an engineer has already decided what needs balancing. That handoff carries the promise from a planning study into the circuit serving the next neighborhood.

Deeper Analysis 👇 https://devcuration.com/articles/switched-source-raises-over-17m-to-expand-grid-capacity

Sensydia

The physician still has a job inside Sensydia's new billing code. Acoustic signals and AI analysis lead to a report that a clinician reviews, interprets and brings into care.

Sensydia wants pulmonary pressure information in the office where heart-failure treatment decisions happen. A measurement becomes useful through the person responsible for acting on it.

Sensydia announced an oversubscribed initial closing of more than $6M on October 13, led by Colle Capital Partners with support from Orlando Health Ventures. The money is intended to finish the ongoing ACOUSTIC-1 study and prepare its Cardiac Performance System, or CPS, for commercial launch.

CEO Anthony Arnold is carrying that plan with cofounders William Kaiser, CTO, and Aman Mahajan, CMO. Historical cofounder Tom Bruggere helped establish the company; Rusty Page now serves as President and COO. The team brings engineering, clinical judgment and device operations together.

CPS acquires acoustic and electrical signals, then uses machine-learning algorithms to estimate cardiac function. Its clinical promise is access to hemodynamic information without sending every assessment through an invasive procedure. That remains a development goal: CPS is investigational and has not yet received FDA authorization.

The pivotal study compares CPS measurements with right-heart catheterization. Training against that reference is central to the investment case Victoria Grace, Colle Capital Partners' founding partner, describes. A clinician asked to trust an estimate deserves to understand what the estimate was measured against.

Category III CPT code 1036T took effect on July 1, 2026, covering passive acoustic and electrical signals, algorithmic analysis and clinician interpretation. Payment remains a payer decision; the code gives this work a defined billing description. The physician's interpretation remains part of the work.

Orlando Health Ventures led Sensydia's approximately $8M financing in May 2023 and is supporting the company again. The change since that earlier round is concrete: a pivotal study is underway, a dedicated code exists, and the new capital is aimed at completing the clinical and launch preparation around them.

For the clinic, the attraction is a pressure assessment that can fit beside a treatment conversation. Sensydia's team is working toward the point where a physician can bring that information into the visit and carry responsibility for what it means for the patient.

Deeper Analysis 👇 https://devcuration.com/articles/sensydia-raises-over-6m-to-advance-cardiac-assessment

Soufflé Therapeutics

Juvia Group arrived at Soufflé Therapeutics with a relationship older than the firm. CEO Amir Nashat says Stefan Vitorovic had already helped think through its pipeline, people and fundraising, while Sujal Shah joined its board 2 years ago. Juvia itself was founded in 2026 by investors and operators.

That history now has a $135M Series B behind it. The Boston-area RNA medicines company announced the oversubscribed close on October 13, led by Juvia, with TCG Crossover Management, BB Biotech, Fidelity Management & Research Company, Balyasny Asset Management and an unnamed sovereign wealth fund joining the syndicate.

Returning backers include Amgen Ventures, ARCH Venture Partners, Bessemer Venture Partners, Breyer Capital, The Invus Group, Janus Henderson, Leaps by Bayer, Polaris Partners, Safar Partners, TAS Partners, Temasek and Vida Ventures. A long investor roster has a fairly specific destination: clinical studies for 3 lead programs and further preclinical work.

Soufflé builds around the address on a cell. Its REMEDY engine identifies receptors and develops binders designed to carry RNA medicines into the tissue where they need to work, extending delivery beyond the liver. A therapy's destination becomes part of its design.

Founder and CEO Amir Nashat and Founding CTO Vadim Dudkin lead that effort alongside CBO and Founding President Susan Langer. The scientific founders include Victor Kotelianski, Daniel Anderson, Robert Langer, Marino Zerial and Brad Pentelute, bringing delivery and genetic-medicine expertise into the same company.

The first program addresses facioscapulohumeral muscular dystrophy, or FSHD. The second targets the PLN gene in heart muscle cells for heart failure with reduced ejection fraction; a third pursues an undisclosed muscle and metabolic target. Soufflé's current website says clinical-trial enrollment has begun in 2026. Patient results still have their own timetable.

Bayer entered a heart-targeted siRNA collaboration in January, with financial terms undisclosed. Meanwhile, WSJ reported just under $200M in combined seed and Series A funding last October. Partnership economics and equity rounds belong in separate accounts.

The interesting portfolio decision is how widely to carry the delivery engine while each program develops its own clinical evidence. Nashat now has more institutions around that decision, and the patients entering studies will bring a kind of information no financing meeting can supply.

Deeper Analysis 👇

https://devcuration.com/articles/souffle-therapeutics-raises-135m-series-b

Godspeed Capital / GreenVest

Brett Berkley is taking over a company that puts its own money ahead of the shovel. GreenVest, LLC's new CEO inherits a restoration business built around finding land, financing the work and carrying responsibility for the result.

Godspeed Capital Management has acquired GreenVest, establishing its new environmental platform with a company serving public and private customers across the Mid-Atlantic and Northeast. Financial terms were undisclosed. Brett Berkley moves into the CEO role from SVP; founder Doug Lashley is welcoming a partner to expand the business he built.

The customer can be an agency managing stormwater obligations or a developer needing wetland and stream mitigation. Their purchase involves a piece of geography, a regulatory process and living systems expected to function after construction crews leave. A tidy drawing has limited comfort value when those obligations still sit on your desk.

GreenVest brings land acquisition, ecological design, permitting, construction, monitoring and long-term stewardship into the delivery process. Its fixed-price model is designed to give the buyer a defined cost while the provider takes on financing and performance responsibilities. The buyer is trusting somebody else to keep a complicated project together.

Godspeed sees a platform it can expand through organic growth, investment in people and capabilities, and strategic acquisitions. Partner James Reid points to a fragmented environmental-services market; Godspeed cites watershed protection and resilience demand. Those are their investment judgments, with growth plans still ahead of the team.

For an operator, the financing matters because the sales promise reaches well beyond signing the contract. A provider funding work upfront needs room to manage the interval between land access, approvals, field execution and payment. Expertise has to survive that interval too, through decisions made by scientists, engineers, attorneys and people standing on the site.

Expansion broadens Brett Berkley's job. A new watershed brings new landowners and local conditions. Buying a team can bring experience into the company; turning that experience into reliable delivery requires keeping the people and their judgment close to the work.

GreenVest's next geography will arrive with its own relationships to earn, even as Godspeed helps put more resources behind the restoration promise.

Deeper Analysis 👇 https://devcuration.com/articles/godspeed-acquires-greenvest-environmental-restoration

Harbor

An agent's execution trace could become part of an insurance negotiation. Harbor is building toward that commercial handoff, where the record of what software can do starts to shape who will carry its financial consequences.

CEO Jesus Santos and co-founders Daniel Elechiguerra, Cristian Bartolome and Luis Botin met at Stanford. On October 13, Harbor emerged from stealth with $10M led by Base10, Abstract and 8VC, alongside Restive, NFX, Innovation Endeavors, Kibo Ventures, Masia and MVP Ventures.

Santos previously worked at Crossmint, the agentic-payments company. That background gives this story a useful setting: an agent with access to a payment tool can turn a software decision into money leaving an account. Somebody has to define the permission, examine the action and account for a mistake.

Harbor describes a platform that inventories agents, measures operational, security and regulatory risks, and checks decisions against company rules. Monitoring, static analysis and adversarial testing feed that assessment. Runtime guardrails are designed to intervene before an unauthorized action reaches execution.

The useful ambition is to give those findings an underwriting job. A record of an agent's tools, permissions and failures could help an insurer understand the exposure, while controls give the customer a way to change it. The same evidence would have to make sense to the engineer maintaining the agent and the institution considering its liability.

There is an insurance business inside the software story. Harbor's legal disclosure identifies Miralta Q, Inc., doing business as Harbor, as an agent and surplus-lines broker. Carriers issue the coverage, with availability and protection governed by jurisdiction and actual policy terms. No blanket promise that every AI mistake gets paid belongs in that description.

Harbor says its first customers are banks and insurers in the US, Europe and Latin America. Named customers, revenue and production-scale results are undisclosed, so the announcement gives us a clear market focus without a scoreboard. The founders are building for organizations where permission to automate has to survive questions from engineering, risk and insurance teams.

An agent changes when someone gives it another tool or broader access. Harbor's proposed relationship with that customer has to keep following those changes, carrying the technical record into the next conversation about what the institution is prepared to let the agent do.

Deeper Analysis 👇 https://devcuration.com/articles/harbor-raises-10m-seed-for-ai-agent-risk-and-insurance

Brellium

A payer enforced a documentation rule before it appeared in the public bulletin. In Zach Rosen's July account, a Brellium customer learned about the requirement through an audit and a repayment demand.

Try putting that sequence in a clinician's calendar. Deliver the care, write the note, discover the missing requirement later. Someone still has to explain what to change before the next patient arrives.

Brellium is taking that work into more specialties with $23M in venture funding. Left Lane Capital and First Round Capital led the financing, with Digital Health Venture Partners participating. The October 13 announcement also introduced CortexLM, its system for reviewing clinical notes and proposing corrections before billing.

Co-Founder and CEO Zach Rosen built the New York company with Henry Kasa, now Co-Founder and COO, Alex Le-Tu, and Max Katzman, Co-Founder and Head of Engineering. Their product sits where clinical judgment, documentation and payer expectations meet, with the provider responsible for making those pieces agree.

The software reviews notes against the requirements that apply to a provider. Finding an issue early gives the team a chance to correct the record while the visit is still close enough to remember. That timing matters to the people who would otherwise spend their day reconstructing yesterday's work.

Rosen described another consequence in the July interview: compliance feedback becomes training. A useful flag should help someone understand the next note, too. Brellium can sell a review and become part of the routine through which a clinical team improves its documentation.

For the returning lead investors, the expansion carries a very practical ambition. Brellium plans to enter cardiology, pain management and primary care while going deeper into the revenue cycle. Each new specialty brings another version of the question clinicians already face: what evidence needs to travel with this care so the payment survives review?

There is an organizational job inside that product promise. A clinic needs to know who receives a flag, who decides the correction and how the lesson reaches the rest of the team. More reviewed notes can make that job visible across the practice instead of leaving it inside a small audit sample.

Brellium's next specialty will bring its own clinicians, its own payer requirements and another set of notes waiting to become claims. The founders are funding the work of getting those people into agreement before the bill leaves the clinic.

Deeper Analysis 👇 https://devcuration.com/articles/brellium-raises-23m-cortexlm-specialty-expansion

Workday / C1.ai

A promotion can leave an employee carrying two jobs' worth of permissions. The new title reaches HR; access to the old team's systems can hang around until somebody notices.

Workday wants that change in responsibility to travel farther. On October 13, it signed a definitive agreement to acquire C1 (C1.ai), founded by CEO Alex Bovee and CTO Paul Querna. The intended connection is practical: use workforce context to govern what people and AI agents can reach across the business.

Alex Bovee and Paul Querna met building Zero Trust products at Okta. They started their company in 2020, when access reviews, scattered permissions and manual requests were already wearing out IT and security teams. The enterprise now has software doing work alongside those employees.

C1.ai's current identity-governance product uses HR and directory events to start onboarding, role changes and offboarding. Its policies govern requests, temporary privileges, reviews and removal across connected applications. A change in somebody's department can become an instruction to withdraw old access and establish the new baseline.

That progression had investors behind it. Accel led the $15M Series A in 2022. Felicis led a $12M extension that expanded the same round to $27M in 2023. Greycroft led the $79M Series B in 2025, with CrowdStrike Falcon Fund, Accel, Felicis and others participating. Those are historical financing figures; the acquisition price is undisclosed.

The agent version puts another decision inside the workflow. An agent acting for an employee may need only part of that person's access; an autonomous agent needs an owner and a defined scope of its own. Workday's Agent System of Record already manages agent records. C1.ai is intended to connect those records with permissions across other enterprise systems, including when an assignment ends.

Gerrit Kazmaier, Workday's president, product and technology, describes the combination as a way to govern the mixed workforce. Closing is expected in Workday's fiscal Q4 ending January 31, 2027, subject to customary conditions. The integrated capabilities remain a plan.

HR, IT and security leaders now share more of the purchasing conversation. The people recording a promotion and the people removing yesterday's permissions would be working from a more closely connected account of what that employee is supposed to do. Keeping it current will involve the application owners who have to carry the change through.

Deeper Analysis 👇 https://devcuration.com/articles/workday-to-acquire-c1-ai-identity-governance

IonQ / Elisity

Quantum security eventually has to answer an ordinary question: should this device be allowed to talk to that one? IonQ is putting up to $1.3B in stock behind a company already helping enterprises make that decision.

On October 12, IonQ agreed to acquire Elisity, whose microsegmentation software controls communication among users, devices and workloads. The price includes up to $150M in performance-based earnout opportunity. Closing is expected in Q4 2026, subject to regulatory approval and customary conditions.

Niccolo de Masi, IonQ's chairman and CEO, is bringing quantum-security expertise closer to the people responsible for keeping enterprise networks running. James Winebrenner, Elisity's CEO, will continue leading the business after closing. CTO Michal Garcarz leads its technology work, with cryptographic inventory on the proposed roadmap.

Elisity's IdentityGraph collects identity, asset and risk context, then helps teams simulate access policies before enforcing them through supported network infrastructure. That sequence matters in a hospital or industrial environment. A tighter rule still has to leave the legitimate work enough room to happen.

The proposed combination would pair those controls with IonQ's post-quantum cryptography and quantum key distribution capabilities. The operating opportunity sits between finding cryptography that needs attention and understanding the devices, communications and business processes attached to it.

Elisity has spent years earning that position. Two Bear Capital and AllegisCyber Capital co-led its $26M Series A in 2021; Insight Partners led a $37M Series B in 2024. TCV and Chevron Technology Ventures later added backing. Investor proceeds remain undisclosed.

IonQ reports more than 100 global Elisity deployments and zero churn among deployed customers. These are company-reported; target revenue remains undisclosed. They explain why an established customer relationship belongs beside the technology in this purchase.

NIST already has finalized post-quantum standards available for migration, while the arrival date of a computer capable of breaking today's cryptography remains uncertain. Security teams can inventory dependencies now without putting a speculative date on the calendar.

For Winebrenner's customers, the work ahead runs through the systems they already depend on. As encryption changes, the people deciding what may communicate will still have a hospital, factory or enterprise to keep running.

Deeper Analysis 👇 https://devcuration.com/articles/ionq-to-acquire-elisity-in-up-to-1-3b-quantum-security-deal

Albert Invent

A new ingredient can send a familiar product back to the lab. Albert Invent is building around the question that follows: what have we already learned about making it work?

The Oakland company announced a $55M Series B on October 13, led by TCV. New investors Delta-v Capital, Socium Ventures, HighSage Ventures, Coral Capital and Legion VC join returning backers Coatue, J.P. Morgan Growth Equity Partners, Index Ventures and Applied Ventures.

CEO Nick Talken and chairman Ken Kisner know the work behind the formulation. Their previous company, Molecule, was acquired by Henkel in 2019. Index lists them alongside Neelesh Vaikhary and Zack Kisner as Albert's founders; Zack Kisner is now Chief Work Officer. That is a team with chemical-industry history behind its software ambitions.

A failed experiment still contains something somebody paid to learn. Its value depends on another scientist being able to find the result, understand the conditions and decide whether it belongs in the next formulation. Paper notebooks and disconnected spreadsheets make that inheritance surprisingly difficult.

Albert OS structures materials, projects and experiments into a shared record. Ask Albert works inside that record, helping scientists retrieve prior knowledge, analyze results and develop experiment candidates. Valence handles chemistry-specific retrieval; Breakthrough models help rank possible formulations. The chemist still chooses what deserves a physical test. Adoption puts that history into the hands of people who did not run the original test.

The customer relationships explain the commercial opportunity. AkzoNobel's Digital Workbench partnership reaches an R&D organization with more than 2,000 professionals across 70 labs. Nouryon's BeautyCreations uses Albert to let personal care formulators search its ingredient and formulation knowledge through natural language. The same scientific record can serve a laboratory and the customer trying to build a product from it.

Albert reports 26x growth in project creation and 6x growth in its scientist user base over the past year, with users across 44 countries. Those are company disclosures, without absolute baselines or revenue attached. The $55M will expand its chemistry-native AI capabilities.

For R&D leaders, adopting the software also means helping different teams describe their work well enough to learn from each other. Albert's next formulation suggestion arrives carrying that organizational history, and another chemist has to decide where to take it.

Deeper Analysis 👇 https://devcuration.com/articles/albert-invent-raises-55m-series-b-chemistry-ai

Intake AI

Four days before college, Miles Wingfield's father suffered a serious stroke. A clinical trial was still under consideration, and Wingfield watched the research team spend precious time hunting for documents and requesting system access.

That experience now sits inside Intake AI's product decisions. The company raised a $2.5M pre-seed led by Kapor Capital, with Blankspace Ventures and Act One Ventures participating, to build AI agents for clinical trial study start-up and operations.

Co-founder and COO Miles Wingfield is building it with CEO Samir Hooker and CTO Lakshman Peri. Hooker previously built VitifyHealth around trial access after his mother's health struggles. The founders arrived at this work through people who needed clinical research, then spent time inside the sites responsible for delivering it.

Their starting point is unusually practical: the source document. A protocol explains the study; somebody still has to turn its requirements into the forms a research team will use at each visit. Intake AI builds those documents from the protocol and case-report requirements, then supports the handoff into existing eSource systems or onto paper.

Paper belongs in that sentence. A site can buy useful automation while continuing to work in the format its study requires. Asking a coordinator to rebuild the rest of the office would make the software itself another start-up project.

The company also describes cited answers across study materials and connections to the systems holding a site's information. The question can travel across a protocol, an operating procedure and a budget, instead of sending another person through each file separately.

Human review stays in the workflow. Approved forms move into the site's existing systems, and Intake's browser extension describes checking the values it entered and pausing when a mapping needs attention. The coordinator gets a way to examine the work before it becomes someone else's problem.

Wingfield says Intake AI is already live with research sites across the country. The public announcement leaves deployment counts and financial results undisclosed; its clearest commitment is to keep adding workflows alongside the teams using them.

The next study will bring its own protocol, visit schedule and paperwork. Intake's founders are building toward the moment a coordinator can accept that study without accepting quite so much copying along with it.

Deeper Analysis 👇 https://devcuration.com/articles/intake-ai-raises-2-5m-for-clinical-trial-operations