What the System Has to Notice

Nirva

A dead battery is part of Nirva's design brief. The company wants its jewelry to remain worth wearing even when the AI has gone quiet.

That puts co-founder and CCO Hanying Hu inside the central business question: how do you earn a place on someone's body every morning? Her jewelry background sits alongside CEO Wei Lyu and CTO Jason (Qianyuan) Chen, co-founders who previously worked at Meta Reality Labs. The engineering and the taste have to persuade the same customer.

Nirva announced a $25M pre-Series A on October 22, taking total funding to $33M. The earlier $8M seed was led by Lightspeed with South Park Commons backing. Neither a lead investor for the new round nor a valuation has been established publicly, so the investor row lists only the earlier Seed backers.

The San Francisco team is building an AI necklace or bracelet for personal reflection. Nirva describes a product that focuses on the wearer's speech and turns daily context into a journal, observations about moods and relationships, and guidance through a companion app. The purchase also asks for trust in what the device chooses to notice. Understanding the person over time is the ambition. That requires the person to keep showing up wearing it.

Jewelry design is doing commercial work here. An app can wait in a pocket while its owner gets on with the day. A pendant has to fit the clothes, the occasion and the person's willingness to have an AI device present. Hanying Hu's contribution reaches into how much useful context the software ever gets to work with.

South Park Commons' early account makes the development less abstract: its prototype needed work on software, battery life and appearance. Those are connected problems for this product. A better reflection tomorrow depends on a wearable somebody was comfortable keeping on today, and on an interpretation that makes returning worthwhile.

Nirva reports a 9,000-person waitlist and a team of 20. CEO and co-founder Wei Lyu has put the expected price around $249, with winter orders and first shipments targeted before year-end. Interested people are still prospective buyers; their daily habits will be formed after the product arrives.

The open roles span audio, AI memory, recommendations, research, design and growth. That is a substantial range of work behind a small piece of metal. Nirva is preparing to carry a conversation from someone's ordinary day into a private reflection they may choose to read, trust and act on.

Pathos AI

Bristol Myers Squibb is putting its own research decisions closer to the software it just helped finance. At Pathos AI, the lead investor is also becoming a user of the drug-development engine.

Pathos closed a $237M Series E on October 23 at an approximately $3B post-money valuation, led by Bristol Myers Squibb with existing investors participating. CEO Iker Huerga now has fresh capital for a company that develops its own therapies while extending its AI platform into another drugmaker's portfolio.

The financing follows a $365M Series D in May 2025, giving this latest raise a substantial prior capital base to build on.

Co-founders Eric Lefkofsky and Ryan Fukushima started Pathos around the connection between patient data and drug development. The New York company has built Foundry to keep reassessing which assets to pursue, which patients to study and where a program should change direction as evidence accumulates.

That job reaches beyond selecting a promising molecule. Pathos describes an operating model in which Scout ranks opportunities and small Sprint teams advance individual programs, with laboratory work feeding back into the shared AI core. The clinical team still has a drug to develop while the engineers have a system to improve.

The company reports four active oncology assets. Funding priorities include moving JSKN016, a TROP2/HER3 bispecific antibody-drug conjugate, toward registrational studies, alongside development of pocenbrodib, AZD4241 and DO-2. Those are separate programs with different biological questions drawing on the same technology investment.

Bristol Myers Squibb gets non-exclusive Foundry access across its broader portfolio and three-year exclusive rights for neuroscience. That arrangement gives Pathos room to serve a large research organization while defining a field where access is reserved. The commercial terms matter to the shape of the platform business.

Tempus AI supplies another essential piece. Its filing describes a five-year neurology-data agreement under which Pathos will pay up to $55M, depending on delivery. The planned records support work beginning with Alzheimer's and other neurological conditions. Data access has a budget and a calendar of its own.

Huerga is taking a platform built around oncology into that new relationship while keeping the cancer pipeline moving. BMS researchers will be asking their own questions of Foundry, and Pathos's program teams will keep deciding which patients, trials and assets deserve the next commitment.

PataBid

An electrical estimate has a supplier account hiding inside it. Count the right materials, then price them for the contractor who will actually buy them. That second job gives PataBid's funding story its commercial weight.

Founder and CEO Melvin Newman spent more than 15 years in construction estimating. PataBid began in 2017, with Paul Burris also a co-founder. Quantify followed: software that connects drawing takeoff, material pricing and ordering for electrical, mechanical, plumbing and fire-protection contractors.

PataBid announced a $2M pre-seed round on October 20, led by Graphite Ventures, with Sprout Fund, Bluesky Equities, Inverted Ventures and previous angel investors participating. The company leaves the dollar denomination unspecified, so that figure stays as announced.

Quantify's supplier integrations make the product more interesting than a faster counting exercise. Contractors can bring their own City Electric Supply or Gescan account into the estimate, import current material prices and order through the same platform. The purchasing relationship gets a place beside the drawing.

That matters to an estimator deciding what a job should cost. A material count still needs a price attached to it, and the price needs to belong to the business submitting the bid. Moving those steps together gives the supplier partnership a role in the work itself.

The company reported more than 200 contractors using Quantify in its financing announcement. That is company-reported adoption. It gives the new investors an existing workflow to build around while leaving revenue, retention and profitability outside the public evidence.

Newton is the next development effort: a conversational estimating agent for questions about drawings and specifications, project risks and material or labor scenarios. It is in beta with selected customers, with commercial launch planned for 2027. Quantify is already available.

The proceeds are intended for AI development, supplier partnerships, distribution and North American commercialization. Those priorities fit together. A contractor considering new estimating software has to bring its drawings, pricing relationships and habits along with it.

Graphite is backing a company whose expansion depends partly on getting more of that supplier network into the estimate. For Newman, each additional connection has to make sense to the person preparing a price before the crew ever arrives.

Tractian

JP Voltani is building for the moment an electrical problem starts looking like a mechanical one. The CTO at Tractian describes machines whose failures cross disciplines, even when the tools watching them do not.

That puts a practical demand on industrial AI: give the maintenance crew enough context to decide which repair belongs on the schedule. A warning needs a route into the work, especially when the person who knows that machine best is no longer on the shift.

Tractian announced a $130M Series D on October 22 at a $1.4B valuation. Sapphire Ventures led, with M&G's Catalyst co-leading. N47, General Catalyst, Bosch Ventures, 53 Stations, Monashees, Endeavor Catalyst, Norte Ventures and Y Combinator also participated. Sapphire led the $120M Series C in December 2024, giving this new check some operating history behind it.

CEO and co-founder Igor Marinelli grew up watching his father maintain equipment at a paper mill. Fellow founders Leonardo Vieira, CCO, and Gabriel Lima, COO, now help run a business that connects its own sensors, AI models and maintenance software. The founding problem has acquired a much larger deployment map.

The platform combines vibration, ultrasound and other machine signals to identify likely failure causes and recommend action. Tractian also says it runs equipment to failure inside its AI Center. Each controlled breakdown supplies observations a customer would rather never collect on a working production line.

The company reports deployment across more than 2,000 critical facilities. Its announcement names Eli Lilly, McKesson, Whirlpool, Coca-Cola bottling operations, Cummins and Embraer among its customers. Those are company-reported relationships, with very different machines and operating conditions behind the logos.

For a plant manager, the purchase reaches beyond an earlier alarm. It is a way to organize maintenance while production, parts and people are still available to plan around. Tractian's decision to own hardware as well as software gives it a direct route to the evidence used in that planning.

The new capital will scale multimodal asset intelligence worldwide. Sapphire points to further European expansion and opportunities in data centers, defense and energy infrastructure. That brings Tractian closer to the machinery supporting the AI buildout itself, where a cooling-system repair has a place in the compute budget long before anyone orders another GPU.

Flapping Airplanes

Ben Spector is building an AI lab where a new idea about learning can reach all the way down to the GPU. At Flapping Airplanes, the researchers get a say in the machinery that decides which experiments are practical.

That choice now has a $1B Series A behind it. Index Ventures and Lux Capital co-led the round announced October 21, with Sequoia, GV, Kleiner Perkins and other backers supporting the lab. The reported post-money valuation is $6B.

CEO and co-founder Ben Spector, co-founder Asher Spector and co-founder Aidan Smith are working on a problem that survives every new chip shipment: getting more useful intelligence from the resources already available. The company announced $180M at its January launch.

Their approach reaches into the software running on the silicon. Flapping Airplanes says its megakernel framework can combine training and inference inside a single GPU kernel, overlapping work that waits on memory with work that keeps the math units busy.

That matters beyond a faster run. The company says the same framework lets researchers explore different architectures, learning objectives and optimizers. When the system underneath an experiment becomes more flexible, the researcher gets more room to change the experiment itself.

Ben Spector has worked on this territory before. His Stanford research includes megakernels, with public code and workload-specific results. That history gives the systems ambition substance; the lab's broader claims about learning efficiency still need their own evidence.

Index describes about 24 people working in person in San Francisco across kernels, pretraining, post-training and inference. It also describes a hire who had never trained a model before taking responsibility for a pretraining run. A small team is being given unusually broad technical responsibility.

For robotics and scientific discovery, two areas the company names, the attraction is learning from examples that are difficult to collect. Better use of a GPU and better use of data solve different parts of that problem. Flapping Airplanes is putting them inside the same research organization.

Index's Mark Xu is joining the board as that organization grows. The team is choosing how research gets done as well as which ideas deserve it. The capital gives the founders more scope to recruit people who can move between an algorithm and its implementation, carrying a promising learning idea far enough into the machine to find out what it can do.

Star Therapeutics

Sandip Panicker has already helped turn an antibody idea into an approved medicine. At Star Therapeutics, the scientific co-founder is back at the part of the job where the next molecule still has to earn its evidence.

That is a useful place to start with Star's $200M private financing, announced October 23. OrbiMed and Janus Henderson Investors co-led the round to advance preclinical programs into the clinic and add new programs to the pipeline.

The South San Francisco company has just lived through the other end of that journey. Incyte completed its purchase of Star's wholly owned subsidiary Vega Therapeutics in July, taking the Phase 3 von Willebrand disease program VGA039, also called latarcibart, into its hematology portfolio.

The price was $1.25B upfront, with up to $750M tied to sales milestones. Those contingent dollars belong in the deal terms, not in a measure of what the sale has already paid.

Founder and CEO Adam Rosenthal built Star in 2018. Panicker, co-founder and CSO, previously led discovery biology at True North Therapeutics, where his work helped produce Enjaymo, the approved treatment for cold agglutinin disease. The team brings that experience back to choosing targets, designing antibodies and deciding which early findings deserve another experiment.

Investors are financing those decisions before Star has disclosed the identities of its next programs. A buyer has taken responsibility for the advanced asset, while the people who built the discovery effort are being backed to develop more candidates.

Janus Henderson was also a participant in Star's $125M Series D in 2025, a round co-led by Sanofi Ventures and Viking Global Investors. The new financing arrives with a different portfolio of work attached, and the financing amount says little on its own about what the company now owns.

Agent Capital, Catalio Capital Management, Cormorant Asset Management, New Leaf Venture Partners, NYBC Ventures, RA Capital Management, Redmile, Sofinnova Investments, Soleus Capital and an unnamed sovereign wealth fund also participated.

Star describes natural-history studies and patient-advocacy partnerships as part of its research approach. Understanding the disease helps determine what a useful antibody should accomplish.

For a biotech team, a successful sale can send the company back toward discovery. Star now has another funded opportunity to work through the biology, select its candidates and bring a new group of molecules into clinical development.

PreScalar

A part can get older while sitting perfectly still. Change its firmware, and yesterday's finished component can become today's reprogramming job.

That is the factory problem PreScalar is taking on with its $7M Seed round, announced October 22. Dynamo Ventures and BMW i Ventures co-led, with Blue Wire Capital, Uplift Ventures and Antler participating. The San Francisco company is bringing testing, software deployment and production data into the same workflow.

CEO Skyler Salman, CTO Benjamin Fredrickson and COO John Cathcart founded PreScalar in 2025 after building hardware and production systems inside demanding engineering organizations. Salman led testing at Cruise, Rivian, Waymo and Google. Fredrickson's background includes safety-critical robotics software; Cathcart's includes getting vehicle programs into production. They know how much engineering disappears into the equipment behind the equipment.

A software team can release an update while a batch of components is already in inventory. Somebody then has to find the affected units, load the right version, run the right tests and preserve a record of what happened. Manual reprogramming or shipping parts back makes a code change somebody else's logistics problem.

PreScalar connects those jobs through modular test equipment and software. The company describes a system that manages firmware, executes tests and captures device-level data across production. A manufacturer can begin with one station, then extend the same approach across a line.

The adoption details deserve attention. PreScalar also supports running existing test scripts on customer hardware. A manufacturing team has accumulated knowledge inside those scripts; keeping that work usable gives an outside platform a more practical route onto the floor.

BMW i Ventures puts the investment in the context of increasingly software-defined products. For the buyer, the conversation gets concrete quickly: which revision reached this unit, what passed, and where is the record when engineering needs it? A shared dataset can make those questions part of routine production work.

The capital will support R&D, hardware manufacturing, hiring and commercial deployments. PreScalar says it is already working with manufacturers, without naming customers in the announcement. Its founders are now taking responsibility for infrastructure other teams can use as firmware keeps changing around the physical batch waiting to ship.

Nevis

Focus Financial Partners considered building its own AI platform. It also considered assembling point solutions, then chose Nevis to connect the work across its advisory business.

That is a consequential purchase. A wealth manager buying this kind of software is deciding who will help turn a client conversation into the work somebody still has to finish afterward.

CEO Mark Swan, CPO Philipp Burda and COO Ivan Chalov founded Nevis after working together at Revolut. Their bet puts human advisors at the center, with AI taking on more of the administration around them.

Nevis has announced a $70M Series B led by EQT Ventures, with Charles Schwab and QuantumLight joining Sequoia Capital, ICONIQ and Ribbit Capital. Reported total funding reaches $110M, following the $35M Series A announced in December 2025 within a $40M cumulative total.

EQT's Carolina Brochado joins Nevis's board, while partner Kaushik Subramanian becomes its new board observer.

The platform brings client data, meeting intelligence, financial planning and custodial workflows together. An account opening involves collecting information, preparing forms, getting signatures and connecting with the custodian. A polished meeting summary leaves plenty of that job waiting on somebody's desk.

Schwab's arrival makes the relationship interesting. A provider of custodial infrastructure is investing in software meant to carry more of the advisor's work through that infrastructure. The company says the funding will expand agent capabilities, deepen integrations and grow its product and go-to-market teams.

Nevis reports more than 100,000 operational tasks completed, up to 10 hours reclaimed per advisor each week and 30% more prospect meetings across deployments. Those are company-reported outcomes. The reported $700B in customer assets describes the firms using its platform, not Nevis revenue or money the startup itself manages.

Focus's deployment also puts people back into the AI story. Nevis describes strategists and engineers working with the customer's operations and experience leaders to configure the platform around the firm. That work belongs in the buying decision alongside the demonstration.

For an advisory business, an extra client meeting has a chain of obligations attached. Somebody has to prepare it, remember the details and carry out the follow-up. As Nevis expands its reach into those steps, its customers are bringing a software supplier closer to the service their own names stand behind.

DropGenie

A patient sample has a spending limit. At DropGenie, the question is how many useful experiments a scientist can get out of it before the material runs out.

Alison Hirukawa, co-founder and CEO, is building around that constraint with co-founders Hugo Sinha and Philippe Vo. Their Montréal and Boston company miniaturizes gene-editing workflows so researchers can work directly with precious primary patient cells. The cells carry information the lab cannot simply order again.

BDC Thrive Venture Fund has joined DropGenie's Seed+ financing, bringing the equity round to $5M USD, according to BetaKit. Skeleton Key led the financing, with Merck Global Health Innovation Fund, CQDM, Real Ventures and Anges Québec also involved. The October 26 update adds BDC to a round first announced in June; $5M is the round total, with BDC's individual investment undisclosed.

The engineering shrinks the experiment onto a microfluidic chip. Electrical control moves tiny droplets, and electroporation helps deliver editing material into cells. DropGenie says the workflow can use up to 100 times fewer cells than conventional approaches, while reducing reagent consumption. For a team holding a finite patient sample, conservation changes the menu of questions it can ask.

There is published work behind that proposition. A 2025 Scientific Reports study describes 48 programmable reaction sites and experiments using as few as 3,000 primary human cells per condition. The paper includes DropGenie researchers, so the result belongs to those experimental conditions. It gives a scientist something concrete to examine before discussing a broader workflow.

The commercial decision is just as interesting. DropGenie is offering customers three routes: run protocols in their own labs, have studies performed as a service, or build standardized functional datasets through a data partnership. That lets the buyer decide how much of the machinery and experimental work to own.

For a drug-discovery team, those are different budgets and relationships. An instrument needs a place in the lab. A service needs a study worth outsourcing. A data partnership needs agreement on which biological questions deserve repeated, comparable experiments. AI's appetite for data reaches all the way back to that choice.

The financing supports commercialization as Hirukawa's team prepares a broader rollout in 2027. The next customer can arrive asking for equipment, an experiment or a dataset. DropGenie is preparing to answer all three.

Arcanus Aerial Systems

Yazan Al Homsi bought into the work that happens between a drone design and a Canadian production line. His stated attraction to Arcanus Aerial Systems Inc. was the manufacturing groundwork: licensing arrangements, a production partner and aircraft aimed at civilian as well as defense customers.

The UAE-based investor announced a US$500K investment on October 23. He remains an investor, with no operating, advisory or board role. The round type and other terms were undisclosed, leaving the industrial story to carry its own weight.

Toronto-based Arcanus is assembling its manufacturing capability across several relationships. An April binding term sheet with Portugal's Beyond Vision set out Canadian manufacturing and commercialization rights, with definitive agreements still to be negotiated under that document. A five-year agreement announced in June appointed FeralX Systems as a non-exclusive Canadian manufacturer. Production follows purchase orders; the agreement guarantees no minimum volume.

That is a practical reading of sovereignty. A country needs somebody who can turn access to technology into equipment, then stay responsible when the equipment needs attention.

CEO Rahim Rajwani is leading a business adding that responsibility in layers. On October 21, Arcanus named Pedram Nowroozi CTO and agreed to acquire Infinit Drones and Nested Technologies, the companies he founded. Their engineering, manufacturing, flight-test, training and repair operation in Woodbridge gives the transaction an unusually tangible center. Closing remains pending, with a joint venture already effective.

The customer story reaches beyond defense procurement. Arcanus and Germany's NeuralDrones announced a five-year exclusive partnership to integrate aircraft into wildfire detection and post-fire monitoring. A crew watching for re-ignition needs a useful flow of information while people and aircraft are being committed elsewhere. That puts endurance, integration and support in the same purchasing conversation.

Arcanus also has an LOI for a phased expansion at Oro Station's Defence Campus. That is a planned industrial footprint, and the US$500K announcement supplies no allocation to it. The value of the investor's explanation is its specificity: he pointed to the arrangements behind the aircraft.

For the team joining Arcanus through the proposed acquisition, those arrangements now meet the hands-on work of preparing systems, training the people flying them and keeping a Canadian customer supported after delivery.

DigitalXForce

A supplier questionnaire starts aging the moment somebody signs it. DigitalXForce wants the people approving that supplier to see what changes afterward, while there is still time to do something useful with the answer.

That is the operating problem behind its growth equity investment from Arsenal Growth Equity, announced October 21. The amount remains undisclosed. Founder and CEO Lalit Ahluwalia is taking the North Texas company into a wider enterprise market with capital earmarked for product engineering, global sales and its partner network.

DigitalXForce calls its approach TRiSCM: Trust, Risk, Security and Compliance Management. Underneath the acronym is a practical argument. The person answering an auditor and the person watching a security control should be working from evidence that describes the same business at the same time. A customer buying that connection is also buying a way to trace a board-level answer back to the system that supplied it.

The company says its platform connects 15 modules, more than 250 technology integrations and more than 50 compliance frameworks. Its third-party risk tools combine document review with external exposure monitoring. Internal controls, supplier risk and AI governance feed a shared view rather than another round of requests for yesterday's attachments.

AI JedAI analyzes control evidence and recommends remediation. XForce GPT produces risk narratives and board reports. X-ROC, the XForce Risk Operations Center, takes that evidence into monitoring, prioritization and escalation. A failed control has somewhere to go after it appears on a dashboard.

Arsenal General Partner Jason Rottenberg points to the company's technology and international momentum. The fit has some specificity: Arsenal invests in B2B software, including cybersecurity and compliance, where operating work gives the software its recurring value.

DigitalXForce reports 514% revenue growth over three years. Its September funding disclosure already described institutional backing; October identifies Arsenal. Neither disclosure supplies a new check size, so the growth figure gets no imaginary financing total attached to it.

The expansion plan covers North America, the Middle East, Europe and Asia-Pacific. Systems integrators, consultants and regional distributors are part of it. For Ahluwalia, more distribution also means more people depending on the evidence arriving intact, with enough context to decide whose job the repair becomes.

Kivo Health

Sending a tablet with its own cellular connection is a fairly blunt verdict on "download our app." Kivo Health is building for people whose lung care cannot wait for the home Wi-Fi to cooperate.

The company mails patients the tablet, an oxygen monitor and exercise bands, then brings pulmonary rehabilitation into the home. For someone managing COPD, the trip to treatment can become another task that requires the energy treatment is supposed to help restore.

Co-founder and CEO Victor Sadauskas completed medical residency at Stanford. Co-founder and CTO Vaughn Koch previously founded Movity, which was acquired by Trulia. They have put clinical care and the machinery for reaching patients inside the same business.

Axios reported Kivo's $10M Series A on October 23. Allumia Ventures led, with Y Combinator, Corewell Health Ventures, Heuristic Capital and MassMutual Ventures participating. Kivo's current YC profile also confirms the round. The financing arrives with relationships already in place: Stanford Health Care, Priority Health and Optum Arcadia have announced collaborations with Kivo.

That distribution matters because a referral still needs somewhere practical to go. Kivo's current program runs for nine weeks, with an individual assessment, live small-group sessions and ongoing support. Licensed professionals supervise the care; the company describes AI supporting coordination and outreach. The clinician stays in the picture. Referrals pass through medical review, and the referring practice receives an outcomes report. The handoff has to work in both directions for care to continue.

There is evidence worth reading carefully. A peer-reviewed retrospective study of 167 program completers found improvements in symptoms and functional exercise capacity. The selected cohort and historical protocol matter when interpreting those results. Kivo's projected 3.1x return for health plans belongs in a different evidence column from measured clinical outcomes.

For a health system, this creates a route from the pulmonologist's recommendation to a session at home. For a payer, it creates a relationship with a care company whose economics depend on sustained participation. Those are related buying decisions, with different people asking to see the receipts.

Kivo's kit takes connectivity off the patient's list. The care team still has to earn the next appointment, then help that patient keep coming back after the equipment has become part of the furniture.