The Story Silence Writes

FIZE Medical

Hospitals do not buy a critical-care device because the demo went well. They buy after nurses can use it at 3 a.m., data reaches the electronic medical record, training repeats, and somebody stays close enough to solve the problems that appear after installation.

That makes Asahi Kasei Medical's role in FIZE Medical's Series B more useful than the headline number. The Japanese medical-device company was already FIZE Medical's exclusive distributor in Japan. It will distribute FIZE kUO across Europe while co-leading the financing with Rapha Capital Management.

A distributor crossing onto the investor side of the table has seen a different company than a financial backer sees in a pitch deck. It has watched procurement, localization, clinical education, support, and the slow work of earning trust inside critical care. Asahi Kasei Medical is choosing to deepen that exposure.

FIZE Medical announced a first close on September 17 toward a $20M Series B target. The company did not disclose how much closed in this tranche, so the accurate number is $20M targeted, not $20M raised. Rapha Capital Management is returning after helping lead the company's Series A financing, which reached a company-reported $36M after a $6M warrant exercise in March.

Founder and CTO Noam Levine began with a measurement problem: urine output is a critical bedside signal, yet much of the workflow has depended on manual readings. CEO Dror Zerem is guiding the next transition. FIZE kUO continuously digitizes urine output and can send that data into hospital records; FIZE Optima is being developed for predictive, AI-guided decision support.

A monitoring system reports the patient signal more frequently. A predictive platform asks clinicians to trust software closer to the treatment decision, raising the burden for clinical evidence, regulatory clarity, integration, and workflow design.

The capital is intended to advance FIZE Optima and strengthen commercialization. FIZE Medical received CE Mark and UKCA certification for kUO in July, and its Japan launch with Asahi Kasei Medical began in December 2025. Europe now tests whether one strategic relationship can carry investment conviction, distribution reach, and hospital execution at the same time.

The first close gives FIZE Medical room to build while the rest of the round remains open. The more consequential close will happen at the bedside, when continuous data moves from being available to being trusted enough to change what a care team sees and when it acts.

Deeper Analysis 👇 https://devcuration.com/articles/fize-medical-series-b-first-close-fluid-management

Raindrop

A pull request can pass code review and still change what an AI agent does to a real person.

The model changed. A prompt moved. A tool gained permission. The software still runs, the tests stay green, and somewhere inside a customer conversation the agent begins doing the wrong thing with complete mechanical confidence.

Zubin Koticha, Ben Hylak, and Alexis Gauba built Raindrop around that gap. The San Francisco company has raised a $35M Series A led by CRV, with existing investors Lightspeed Venture Partners and Y Combinator participating alongside researchers and executives from Anthropic, OpenAI, and Thinking Machines. The round brings total announced funding to $50M after a $15M seed.

Raindrop already monitors production agents for semantic failures that conventional observability can miss: hallucinated answers, tool misuse, loops, and behavior shifts introduced by a new model or harness. Its new product, Raindrop Simulations, moves that evidence into the pull request.

Simulations replays production traffic and existing tests against a proposed agent change, then uses anomaly detection to find behavior the engineering team expected and behavior nobody thought to put in a fixture. The difficult part is recreating the world around an agent when the tools write to databases, issue refunds, search changing repositories, or did not exist in the original trace.

That is where this round earns its meaning. Monitoring tells a team what already hurt a user. Simulation tries to turn the incident into institutional memory before the next change ships.

Raindrop says Vercel, Framer, Clay, Speak, and Fortune 100 enterprises use its products. Those are company-reported signals; revenue, retention, detection accuracy, valuation, and deal terms remain undisclosed.

CRV is backing the move from an alarm layer into release infrastructure. Lightspeed and YC are doubling down on a team that began with production monitoring and now wants every failure to improve the next test.

The commercial test will arrive inside ordinary engineering work. Raindrop has to reproduce customer environments, protect sensitive traces, control false alarms, and make the gate useful enough that teams do not route around it when deadlines get loud.

Zubin Koticha, Ben Hylak, and Alexis Gauba are betting that an agent team's most valuable evaluation set is the history its users already exposed. The $35M buys time to turn that history into a system that remembers before the customer has to.

Deeper Analysis 👇 https://devcuration.com/articles/raindrop-raises-35m-series-a-agent-simulations

Robigo

Field trials turn sustainability claims into yield data, season after season.

Robigo has closed an undisclosed Series A led by Leaps by Bayer, with Illumina Ventures, SVG Thrive, Congruent Ventures, and Endeavor8 participating. The capital will move engineered microbial crop protection toward U.S. commercial launches and expand the pipeline.

Growers do not receive a second planting because a biological product had better intentions. A treatment must control disease reliably, fit existing practices, and make economic sense against chemistries farmers already know.

Andee Wallace, PhD founded the 2021 MIT spinout with Jai Padmakumar and Connor Sweeney. Wallace now leads Robigo as founder and CEO, carrying their wager that living microbes can compete in commercial crop protection.

Robigo's ARGO platform engineers microbes to produce active ingredients that silence essential genes in specific plant pathogens. A one-time seed treatment is designed to keep delivering those ingredients through the growing season.

The company says its soybean program for sudden death syndrome matched commercial chemistries at half the cost. A second program targeting fusarium wilt in lettuce and berries produced 3x the yield of untreated controls. Robigo says ARGO moved programs from design through field testing in under 20 months, versus a three-to-five-year norm.

Those are company-reported development results, not audited commercial outcomes. The amount, valuation, terms, ownership, and board changes were not disclosed. Robigo's 2028 and 2029 launch dates are targets, not finished milestones.

The investor mix still matters. Leaps by Bayer emphasized the repeatable engine behind multiple soil-borne pathogen programs. The round backs a platform whose value rises only if one product can teach the next. Illumina Ventures brings genomics depth, SVG Thrive and Congruent Ventures bring agriculture and climate networks, and Endeavor8 returns after Robigo's 2023 seed extension.

Robigo closed that seed round at $10M. Later $1.25M NSF and $350K Massachusetts Clean Energy Center awards were grants, so they stay outside the venture total.

The Series A now has to carry a living technology through field variability, regulation, manufacturing, distribution, and farm economics. Useful evidence will arrive across seasons, crops, disease pressure, launch preparation, and the growers deciding whether an engineered microbe deserves space beside established chemistry.

Deeper Analysis 👇 https://devcuration.com/articles/robigo-series-a-engineered-crop-protection

Tempo Therapeutics

A regenerative scaffold has 2 jobs that pull in opposite directions: stay long enough for tissue to organize, then disappear before the body decides the material is the problem.

Tempo Therapeutics has closed an oversubscribed $14.5M financing led by Galaxy Sirius Partners, with Johnson & Johnson participating through JJDC alongside Gideon Strategic Partners, Mesa Verde Venture Partners, YK Bioventures, and other investors.

The filing adds precision. Tempo's September 11 Form D classifies the offering as debt, showing $14M sold toward a $14.5M total before the company announced the completed financing 5 days later. The terms and valuation were not disclosed.

That money arrives after the company's MAP wound matrix moved from materials science into a 40-patient first-in-human study. The fully synthetic, flowable scaffold is designed to fit irregular surgical spaces, let cells and blood vessels move through it, then gradually resorb as the patient's own tissue takes over.

In the randomized multicenter study, 26 post-Mohs patients received MAP and 14 received a hydrocolloid control. The primary safety endpoint was met. MAP-treated wounds reached a favorable Wound Bed Score about 14 days earlier, and scar scores were significantly better at roughly 6 months, while overall closure time remained similar.

This was an encouraging first human signal, not permission to turn 40 patients into a universal claim about regeneration. Tempo has submitted its lead product through the FDA De Novo pathway. Submission still has to become clearance, manufacturing readiness, physician adoption, and performance outside a controlled study.

The company is arranging leadership around that handoff. Director Eric Richman is now interim CEO. Co-founder and former CEO Westbrook Weaver moves into the CTO role, while co-founder Donald Griffin remains CSO and a director. Steven Sands of Galaxy Sirius Partners becomes board chair, and Erik Oros of Gideon joins the board.

Tempo's 2024 $12M Series A helped bring MAP into clinical testing. The new financing is meant to support potential clearance and commercialization in surgical reconstruction while generating data for regenerative aesthetics.

The platform story can get big quickly: wound healing, reconstruction, aesthetics, and other regenerative applications. Tempo now has to prove that a material designed to hand the work back to the body can survive the handoff from promising science to regulated medicine.

Deeper Analysis 👇 https://devcuration.com/articles/tempo-therapeutics-14-5m-map-regenerative-technology

Nex

A child does not care whether growth is financed with Series E equity or a credit facility. The game has to work, the library has to stay fresh, and the parent has to trust what enters the room.

Nex has raised more than $150M in new equity and debt financing for that operating challenge. The package includes a Series E co-led by Baillie Gifford and BAI Capital, with NBA Investments, Logitech, Medici Capital Partners, and the Raine Group participating, plus a new JPMorgan credit facility. Nex did not disclose the split, so this is not a $150M equity round wearing a shorter headline.

The capital follows company-reported traction that changes the scale of the job: more than 1M Nex Playground units sold, distribution through 7,000+ U.S. retail locations, a subscriber base approaching 1M after growing 7x in 18 months, and a library of 60+ games.

David Lee and Tony Sung helped build Nex around a deceptively hard idea. Kids should be able to move in front of a screen without a controller, while parents get a curated system with local motion processing, no advertising feed, and a camera cover they can see. The device gets into the home once. Play Pass has to keep earning a place there.

That makes this a hardware expansion and a relationship expansion at the same time. Nex is broadening retail distribution, adding content partners, planning Connected Play across households, and pushing into Europe and Asia. Canada, the UK, and Ireland are already live. Germany is planned later this year, with Japan and Korea planned for 2027.

The financing mix matters because physical scale creates obligations long before a subscription renewal arrives. More countries mean more inventory decisions, retail relationships, localization, support, content rights, and trust standards. Equity can finance the category Nex wants to build. Credit adds capacity, but it also gives execution a clock.

The leadership additions make the same point. Jeff Shouger is joining as CFO after finance roles at Niantic and Zynga, while Shouger and Bing Gordon are joining the board. This is the kind of bench a consumer company adds when the question shifts from whether people want the product to whether the organization can keep pace with demand.

Nex has sold the first million boxes. The next chapter lives in the quieter handoffs between hardware, content, subscription value, global distribution, and the parents deciding whether the system still deserves space in the room.

Deeper Analysis 👇 https://devcuration.com/articles/nex-raises-150m-global-active-play

Antfly

An AI agent can turn a retrieval mistake into an operating decision before anyone notices which policy, clause, customer record, or product fact the system missed.

That is the pressure behind Antfly, the Portland company founded by James McDermott and AJ Roetker. Antfly has raised a $2M pre-seed round led by Heavybit, with 8-Bit Capital participating, to build a retrieval engine for AI applications and agents.

The stack underneath rarely is. Teams stitch together vector search, keyword search, graph tooling, embedding and reranking services, synchronization, permissions, and custom code that must survive after the demo becomes a product.

James McDermott, Antfly's CEO, and AJ Roetker, its CTO, have lived close to that problem. Heavybit says the pair previously built data systems at billion-profile scale at Lytics. Antfly is their attempt to turn the retrieval layer from a recurring integration project into one system.

The engine combines keyword, vector, and graph retrieval inside a distributed document database. Chunking, embeddings, reranking, and extraction run inside the platform. Teams can self-host it so data and inference remain in their environment, while Antfly Cloud offers a managed private-instance path.

That architecture matters because agents have changed the cost of a bad result. A search user can question a strange answer. An agent may use retrieved context to route a ticket, update a record, recommend an action, or trigger the next tool. The missing clause does not stay inside a search box.

Heavybit General Partner Joe Ruscio calls this the retrieval gap. The investment thesis is that enterprises need lexical, semantic, and graph retrieval with access controls and deployment ownership in one framework. 8-Bit Capital joined the round.

Visiting Media CTO Eric Sniff said his team moved from maintaining a retrieval pipeline to experimenting with indexes and models within days. Useful evidence, yes, but still company-issued testimony rather than audited traction. Antfly disclosed no revenue, customer count, valuation, terms, or independent benchmark.

The commercial assignment is precise: prove that consolidated retrieval removes more operational pain than it introduces, and that buyers will fund this layer before unreliable context becomes an incident they can name.

The $2M gives Antfly room to build the engine and the managed service. The more consequential work is earning a place between enterprise knowledge and the agents increasingly allowed to act on it.

Deeper Analysis 👇 https://devcuration.com/articles/antfly-raises-2m-ai-retrieval-engine

Phigenics

A water sample captures a moment. A hospital has to manage everything that changes after it: temperature, disinfectant levels, stagnation, construction, occupancy, patient vulnerability, and the corrective action somebody must document.

That continuing obligation is where Phigenics built its business, and where Battery Ventures just bought a majority position.

Battery made a majority-growth investment in the Warrenville, Illinois, water-management company. The amount, valuation, exact ownership percentage, and other terms were not disclosed. Battery calls the investment buyout stage; Phigenics describes it as growth capital for products and new markets.

Bill McCoy, Ashton McCombs, and Jay Reading founded Phigenics in 2004. Mark Crockett, MD, became CEO in August 2024 and will keep running the business with the team.

The investor logic sits in the handoffs. Phigenics designs water-management programs. phiAnalytics centralizes documentation, reports, and alerts. Sensors track disinfectant residual, temperature, and pH. Laboratories test for Legionella and other microbial hazards. Then people turn findings into documented action.

That full loop can make the relationship difficult to remove. CDC calls a comprehensive water-management program the primary strategy for controlling Legionella growth and spread. CMS expects covered facilities to maintain risk assessments, control measures, documentation, and corrective actions. The work keeps moving after a sample comes back.

Phigenics reports more than 250 customer organizations, including hospital systems, VA facilities, and ambulatory surgery centers. The figure is company-reported. Revenue, retention, profitability, and measured health outcomes remain undisclosed. The verified record is a 22-year-old operator moving into institutional majority ownership.

Battery General Partner Zack Smotherman and Principal Max-Julian Kaye will join the board. Battery has invested in water-related technology for more than 20 years. Phigenics brings services, software, equipment, laboratory work, and compliance documentation into one customer relationship.

Healthcare remains the center of gravity, while Phigenics plans to expand further into hospitality and retail. That requires making a complex operating model repeatable across facilities with different plumbing, occupants, risks, and regulatory exposure.

The next chapter lives inside the handoff between a sensor alert and the team that must act before water becomes a patient-safety event.

Deeper Analysis 👇 https://devcuration.com/articles/phigenics-battery-majority-investment-water-safety

TowerNorth

A wireless carrier can approve more capacity before lunch. The site that delivers it may still spend months inside real estate talks, zoning calendars, engineering reviews, construction schedules, and the small matter of persuading everyone involved that the deadline is real.

That gap is where TowerNorth makes its money.

TowerNorth secured additional growth capital from Berkshire Partners and closed a new debt facility. The amount, lender, pricing, and ownership impact were not disclosed. What was disclosed is the choice it creates: build new towers, back development partners, or acquire existing assets when economics and carrier demand line up.

Bert Stern and Josh Delman founded TowerNorth in 2012. Bert Stern now leads the company as CEO, alongside COO Tim Jaki and CFO Josh Mossey, getting wireless sites from a carrier's network plan into the physical world.

A viable site needs real estate, municipal approvals, engineering, utilities, construction, tenant installation, and structural capacity for more carriers. Steel is visible. The coordination that makes it useful is where the calendar, capital, and risk live.

Since Berkshire Partners' 2023 investment, TowerNorth says its portfolio grew from 64 towers to 231. It also reports acquiring 28 towers since the start of 2026 while expanding development capabilities and its pipeline of new sites and colocations.

Those are company-reported figures, not audited operating results. They still change the reading. Berkshire Partners is backing a platform already moving across construction, partnerships, and acquisitions, then asking the team to keep choosing the right lane for each opportunity.

The debt facility adds another layer of discipline. It can help TowerNorth move when a site or portfolio becomes available, but the size, cost, maturity, covenants, and provider remain private. Flexibility still has a price.

Wireless demand supplies the reason to build. Carriers can add capacity through more spectrum, better technology, or denser networks. Spectrum is finite, and local approvals do not accelerate because a spreadsheet says demand is growing.

TowerNorth now has to decide which sites deserve construction capital, which developers deserve a partnership, and which towers are worth buying, then move each one through the local work that turns financing into coverage.

The portfolio can grow on paper the moment a transaction closes. A carrier only feels the value when the next site goes on-air.

Deeper Analysis 👇 https://devcuration.com/articles/towernorth-growth-capital-wireless-infrastructure

Embra AI

Robotics teams can collect a mountain of data and still train on the wrong experience.

The camera angle may be off. The sensor streams may not line up. The demonstration may come from a different robot, environment, or task. A dataset can be expensive and technically clean while teaching very little about the machine that has to act.

Anthony Smith is building Embra AI around that selection problem.

The company announced a $1M pre-seed round on September 17 to develop data infrastructure for robotics and physical AI teams. Embra AI says its backers have professional backgrounds at Boston Dynamics, Agility Robotics, and NVIDIA. It did not name the investors, a lead or the terms.

Prior employment is not corporate participation. Three famous names should not become three invented investors by the time the headline reaches the feed.

Embra AI is trying to make robotics data legible before a team spends compute on it. The platform combines discovery, evaluation, curation, and sourcing across video, sensors, robot state, and actions. Its website describes matching data to a target task, embodiment, environment, and sensor stack.

The contributor side is part of the wager. Embra AI asks providers for task metadata and rights confirmations, evaluates submissions for integrity, synchronization, relevance, and coverage, then describes USDC rewards for accepted data.

The need is easier to verify than Embra AI's scale. Google DeepMind's Open X-Embodiment project pooled more than 1M episodes across 22 robot types. The DROID consortium used 50 collectors and 76,000 trajectories for a broad manipulation dataset. Robotics data becomes infrastructure when context survives collection.

Embra AI reports working with three robotics teams and completing its first curated dataset deliveries. The companies are unnamed, and revenue, dataset volume, repeat usage, and independent benchmarks remain private. The $1M is early capital for proving that evaluation can become a product instead of another custom research chore.

The round will support platform development, stronger evaluation, broader operating-condition coverage, and a larger contributor network. Each expansion creates another judgment call about whether two pieces of robot experience can carry value between machines.

Embra AI now has to turn those judgments into trust on both sides: useful enough for robotics teams to buy, clear enough for contributors to share, and disciplined enough that the data keeps its meaning after it leaves the machine that created it.

Deeper Analysis 👇 https://devcuration.com/articles/embra-ai-raises-1m-physical-ai-data

TigerByte Cyber

Modernization gives an old aircraft new reach, then asks its original wiring to survive a threat model nobody drew when the platform left the factory.

TigerByte Cyber is building for that handoff. Co-founders Sage Secilmis and James Brock are focused on hardware security for mission-critical edge systems, where power, space, and downtime are expensive.

The company has emerged from stealth with a $3M Seed round led by Hale Capital Partners, with Tenon VC participating. Galois CEO Rob Wiltbank is joining the board as part of the deal.

TigerByte's Cyber Protection Suite sits between legacy platforms and the new sensors, networks, and software being added to them. The compact hardware layer is designed to validate data, inspect packets, segment networks, and add post-quantum encryption without rebuilding an aircraft, satellite, vehicle, or drone around a new security architecture.

That sounds less glamorous than replacing the platform. It is also closer to how defense and aviation actually modernize: critical machinery stays in service while capability arrives one subsystem at a time.

TigerByte reports more than $7M in contracts involving DARPA, the U.S. Space Force, the U.S. Navy, and other agencies. Federal SBIR records under Web Sensing, the legacy operation tied to TigerByte, document years of FPGA and system-on-chip work across air, space, and ground platforms.

The company also reports flight heritage from a March 2026 satellite deployment and an August Navy field demonstration. Those are company-reported milestones, not independent performance audits, but they clarify the path from government R&D to a commercial product.

The $3M is earmarked for production scale and expanded U.S. manufacturing. A security component can work in a controlled program and still face a brutal production list: supply-chain discipline, certification, integration, support, and customer trust.

Hale's defense focus and Wiltbank's board seat add institutional help around the move from promising technology to hardware that can be manufactured, installed, and supported without slowing the mission.

Valuation, ownership, financing terms, revenue, and current customer count were not disclosed. The useful signal is the transition now in front of the company.

Legacy fleets absorb new capability while staying operational, and every new connection inherits the responsibility of the old machine around it. TigerByte now has the capital to prove its hardware can carry that responsibility at production scale.

Deeper Analysis 👇 https://devcuration.com/articles/tigerbyte-cyber-raises-3m-secure-edge-systems

Angle Health

Health insurance makes a small business pay for uncertainty before it can pay for care. The renewal arrives, the numbers move, and an owner has to decide whether the next dollar belongs in wages, growth, or coverage employees cannot afford to lose.

Angle Health is financing a different answer. Co-founder and CEO Tylon (Ty) Wang and co-founder and CTO Anirban Gangopadhyay announced a $600M equity transaction at a $2.7B valuation.

The headline contains 2 different transactions. Vitruvian Partners led a $200M Series C that puts new capital into Angle Health. A $400M tender offer provides liquidity to existing holders. That second piece changes ownership; it is not another $400M for operations.

New investor Town Hall Ventures joined Blumberg Capital, Portage, PruVen Capital, and Y Combinator. The deal is expected to close later this month, less than 10 months after a $134M Series B that combined debt and equity.

Angle Health sells level-funded health plans and the machinery required to run them. Benefit Builder gives brokers firm, underwritten quotes from a census in minutes. Quote-to-Card handles implementation. The platform connects enrollment, billing, claims, reporting, member guidance, pharmacy programs, and care navigation.

That integration matters where buying power is weakest. Morgan Health found a median 18% premium increase among small and midsize employers. Angle Health reports median customer renewals of 5% to 7%.

The company also reports more than 5,000 employers across 47 states, nearly $1B in annualized premium-equivalents, 120% year-over-year growth, and 4 consecutive quarters of EBITDA and GAAP net income profitability.

Those are company-reported figures, and premium-equivalents are not revenue. Member count, cash flow, medical loss ratio, and independent audits remain undisclosed.

The $200M in primary capital will support the AI-native platform and more local, condition-specific care options across medications, infusions, outpatient surgery, and radiology. The tender gives earlier holders liquidity; the Series C gives the company a heavier assignment.

Every new employer brings another renewal, provider network, claims pattern, regulatory context, and workforce that will judge the product when somebody needs care.

A $2.7B valuation can be agreed in a financing. The durable number will be the renewal a small employer sees one year later, when better infrastructure has to show up as a benefit people can still afford.

Deeper Analysis 👇 https://devcuration.com/articles/angle-health-600m-financing-small-business-benefits

Penelope Health

A prior-authorization team can have the right procedure, the right clinical case, and the wrong payer rule.

The failure may begin in a policy PDF that changed or buried one requirement. By the time anyone sees it, a billing team has inherited rework and a patient is wondering whether care will be paid for.

Penelope Health is trying to move that moment upstream. The company announced a $100M commitment led by Thoreau, with existing investors BHI - Bertelsmann Healthcare Investments, Twine Ventures, and Seedcamp participating.

The company calls it a partnership and commitment, not a named round. No valuation, security type, ownership, board right, allocation, or funding schedule was disclosed.

Sohum Patel and Mattijs De Paepe built Penelope around a deceptively simple question: will insurance cover this procedure?

Sohum Patel practiced in the U.K. NHS and saw administrative complexity interfere with care. Mattijs De Paepe spent nearly 4 years building machine-learning systems at Via Transportation. They are turning coverage policies, prior-authorization requirements, medical-necessity criteria, and coding rules into data that providers, payers, and software can query.

Penelope says its platform covers policies affecting more than 200M Americans across 15,000+ procedure and drug codes. Its API supports coverage checks, policy comparisons, claim scrubbing, prior-auth preparation, appeals, and payer benchmarking.

The timing has teeth. CMS is requiring faster decisions and specific denial reasons from certain payers in 2026, with several FHIR-based APIs arriving in 2027. The market is moving toward structured exchange while teams still search documents and portals for changing rules.

Matt Holt's Thoreau is building across healthcare compute, rules, and research infrastructure. Its partnerships include Penelope and Ensemble Health Partners. No integration has been announced, but the investment logic is visible: payment operations work better when the rules arrive before the claim.

BHI says it led Penelope's pre-seed earlier this year with Twine and Seedcamp. The new commitment supports more products, insurers, and policy types as the team hires across New York and London.

The founders are chasing a valuable handoff. Clinical intent becomes administrative data, then a coverage decision, and the patient lives with what gets lost between those steps. Penelope now has to make the rule layer trustworthy enough that fewer teams discover the requirement after the work is done.

Deeper Analysis 👇 https://devcuration.com/articles/penelope-health-100m-payer-policy-intelligence

CADDi

A drawing can show what a part became. It rarely explains why an engineer accepted one supplier, rejected another, changed a tolerance, or carried an old decision into the next design. That missing history is where manufacturing AI gets expensive.

CADDi has raised a $114M Series D at a $1.2B valuation to keep building an AI data platform around that problem. The round includes Moore Strategic Ventures, Coreline Ventures, Salesforce Ventures, Woven Capital, and HR Tech Fund, with existing backers Atomico, Globis Capital Partners, and JPS Growth funds returning. CADDi says total equity funding stands at $234M.

Co-founder and CEO Yushiro Kato and co-founder and CTO Aki Kobashi started CADDi in 2017 inside the less glamorous realities of manufacturing: quoting parts, correcting defects, managing suppliers, and inspecting physical output. That experience now sits underneath a broader software bet.

CADDi Explorer searches drawings and operational records. CADDi Agent reasons across that information. Six workflow products extend the foundation into design, cost, production, quoting, reuse, and asset lifecycle work.

The architecture matters because a manufacturer does not need another chatbot summarizing a PDF. It needs software that can connect geometry, supplier history, quality records, ERP data, and the judgment an experienced employee never wrote down.

CADDi says sales are more than doubling year over year, customers use the platform in 22 countries, and more than half of Japan's 100 largest manufacturers are customers. Revenue and customer count remain undisclosed, so the useful signal is the depth of work customers are willing to place inside the system.

The capital will support more products, manufacturing-specific models, North America-centered expansion, and hiring. CADDi is also carrying a less fashionable expense: more than 100 customer-success employees and forward-deployed engineers helping factories change workflows.

Manufacturers protect design and cost data because mistakes become inventory, rework, delays, and warranty claims. Winning the contract requires technical capability. Becoming part of the decision process requires trust, implementation, and enough operational patience to survive the handoff from veteran memory to a shared system.

$114M gives CADDi more room to build the intelligence layer. The value will emerge inside thousands of small decisions that no longer disappear when a drawing closes or an experienced engineer leaves the room.

Deeper Analysis 👇 https://devcuration.com/articles/caddi-raises-114m-for-manufacturing-decision-memory