Grishin Robotics

Researchers developed a helium-filled soft robot that floats through indoor spaces using flapping wings instead of propellers. Its 70-gram, touch-safe body is designed for close interaction with people. In an earlier study, 92% of participants spontaneously touched it. The team’s new ACM DIS 2026 Best Paper explores future floating companions that could follow people, guide breathing, interact with pets or deliver gentle reminders. These remain research prototypes and design concepts - not yet an autonomous consumer product.

India has more than 63 million MSMEs, but commercial insurance still reaches only a tiny slice of them. BimaKavach has raised undisclosed strategic funding from Beyond Capital Ventures to make that gap smaller with AI-native business insurance. The product mechanism is concrete: BimaKavach analyzes submissions, operating data, exposure, and claims history, then connects buyers to more than 20 insurers for quotes and binding that can happen in minutes instead of weeks. The demand signal matters because this is not consumer insurance repackaged for startups. BimaKavach already serves more than 5,000 businesses across 2,400+ pincodes, with coverage spanning liability, cyber, fire, engineering, and other commercial risk products. Quick facts👇 ● founders: Tejas Jain ● total capital raised: undisclosed strategic funding ● HQ: Bengaluru, India ● Investors: Beyond Capital Ventures The bigger shift: commercial insurance in India may move from broker-led paperwork to risk-intelligence infrastructure, especially for businesses too small to be served like enterprises but too exposed to stay underinsured.

Applied AI has plenty of pilots. AltaML is trying to make it something enterprises and governments can actually run. AltaML has received a $6.1 million strategic investment from VerticalScope to keep building AltaForge, its platform for production-grade agentic AI workflows, LLM integrations, and automation. The structure is unusual: an 18-month secured promissory note plus a five-year warrant for preferred shares. It is not a classic priced venture round, but it does say something useful about where enterprise AI demand is moving. VerticalScope was already an AltaML client, using the company to deploy applied AI across its platform and operations. That makes the investment feel less like a thematic bet and more like a customer doubling down on a deployment partner. The interesting layer is AltaForge. AltaML is positioning it for organizations that want agents inside real workflows: governed, observable, and tied to measurable operating outcomes rather than demo-day productivity gains. Quick facts👇 ● founders: Cory Janssen; Nicole Janssen ● total capital raised: at least $6.1M ● HQ: Edmonton, Alberta ● Investors: VerticalScope Will the next applied AI winners be the teams that help institutions ship reliable systems, not just better prototypes?

Dogs hide disease until the window for easy intervention has often closed. Wonderdog has launched with $5 million in pre-seed funding led by WNDR and Maveron to bring at-home blood testing and AI health intelligence into dog care. The Los Angeles and New York company is building around a simple gap: annual vet visits are too sparse for conditions that move quietly. Its membership sends licensed veterinary technicians to the home, runs 40+ biomarkers through Antech, and returns vet-reviewed results plus an AI Health Companion that tracks each dog's baseline over time. The AI edge is not a chatbot wrapper. It is longitudinal context: bloodwork, records, breed, conditions, and owner inputs combined so subtle biomarker shifts can become earlier questions for a veterinarian instead of late-stage surprises. Quick facts👇 ● founders: Jeff Safenowitz ● total capital raised: $5M ● HQ: Los Angeles and New York ● Investors: WNDR; Maveron ● Partners: Antech If pet health moves from reactive visits to preventive monitoring, the winning layer may be the one that turns quiet biomarker drift into timely action.

ESGAgent.ai is turning heavy-industry compliance from annual paperwork into live operations. The Brisbane startup raised A$450K in seed funding led by DNX Ventures, taking total capital raised, including grants and earlier pre-seed funding, to A$1.7M. The company is building AI-native software for climate, emissions, safety, modern slavery, biodiversity, and broader ESG governance reporting. The wedge is not prettier sustainability slides. Heavy industry still runs too much compliance through consultants, spreadsheets, scattered evidence, and once-a-year reporting cycles. ESGAgent.ai wants to compress that work from months to hours and turn it into a single system of record for regulated operators. The early buyer signal matters. Startup Daily and Business News Australia report tier-one miners, food manufacturers, and global engineering consultants among customers, exactly the kind of complex environments where audit trails, data collection, and reporting deadlines collide. Quick facts👇 ● founders: Shan Vahora ● total capital raised: A$1.7M ● HQ: Brisbane, Australia ● Investors: DNX Ventures If this works, the moat is not report generation; it is keeping industrial sites audit-ready year-round.

Retailers have spent years buying email tools, then paying people to make those tools useful. Birdsview.com just raised €2.5 million in seed funding to push a different model: Avys, an AI worker that reads a shop's catalog, decides what each customer should hear next, and sends individualized retention emails without manual flows. Fortino Ventures and Newion backed the round. The timing is notable because Birdsview.com is attacking a very specific pain point in e-commerce: repeat purchases are crucial, but catalog-scale personalization still turns into rules, segments, templates, and agency work. What makes Avys more interesting than another marketing automation layer is the product-intelligence claim. Birdsview.com says Avys maps 1.2 million labeled products across 60 industries and processes 52 million email interactions each month, with customers seeing up to 3x more clicks and 8x more revenue per email. Quick facts👇 ● founders: Ozan Brochwitz-Tuerker; Tim Kuesters ● total capital raised: at least €2.5M ● HQ: Leipzig, Germany ● Investors: Fortino Ventures; Newion The contrarian read: in retention marketing, the next software category may not be a better dashboard. It may be the first credible replacement for the operator behind the dashboard.

Octozi turns clinical trial data operations into a human-in-the-loop AI workflow. The New York company raised $3 million in seed funding led by Surface Ventures, with Remarkable Ventures participating, after an earlier strategic investment from Debiopharm. The wedge is narrow but painful: before a drug can move toward approval, sponsors still spend months cleaning, reconciling, reviewing, and reporting trial data across disconnected systems. Octozi connects to those systems and uses agentic AI plus deterministic clinical algorithms to flag what matters, route review work, and keep study teams in control. That matters because clinical development is not short on dashboards. It is short on trustworthy operational throughput. In one controlled study, Octozi reported roughly 6x faster data-cleaning throughput, error rates dropping from 54.7% to 8.5%, and an economic model showing more than $5 million in potential savings for a representative Phase III oncology trial. Quick facts👇 ● founders: Amit Patel; Matt Purri, PhD ● total capital raised: $3M+ disclosed ● HQ: New York City, New York ● Investors: Surface Ventures; Remarkable Ventures; Debiopharm Contrarian part: regulated pharma AI may first win in the unglamorous back office, not the discovery lab.

Cableteque just raised $6 million and is hiring to turn wire-harness manufacturing's PDF-and-spreadsheet mess into software. The Huntington Beach company closed seed funding backed by 11 Tribes Ventures, PHX Ventures, and Mudita Venture Partners for a digital operating system built around one under-digitized niche: wire harnesses. The market gap is specific. OEM requirements arrive as drawings, annotations, and PDFs, while manufacturers still translate much of that into BOMs, sourcing, quoting, labor estimates, and production steps by hand. Cableteque AI parses those unstructured files into normalized configurations, and Quote Summary pulls supplier pricing and labor estimation into one workflow instead of another Excel handoff. That is not generic AI for manufacturing. It is a translation layer between engineering intent and shop-floor execution in a $100B category, with Cableteque reporting 300% year-to-date growth and quote turnaround reductions of more than 70%. Quick facts👇 ● founders: Arik Vrobel ● total capital raised: $6M ● HQ: Huntington Beach, California ● Investors: 11 Tribes Ventures; PHX Ventures; Mudita Venture Partners The bigger shift: manufacturing software may win by compressing quote-to-production friction, not by adding another dashboard.

AI agents can call tools. Paying for those tools is still built for humans. AIsa has raised $6.5 million in total funding, including a new seed round co-led by Alibaba Group and Tribe Capital, to build transaction infrastructure for autonomous agents. The San Francisco company is not trying to be another agent interface. It is building the layer underneath: a resource gateway plus usage-metered payments so agents and developers can discover, access, and pay for models, APIs, data, compute, SaaS tools, and agent services through one programmable interface. That matters if agents move from answering questions to spending money on behalf of users. Account creation, vendor contracts, subscription plans, API keys, and checkout pages were designed around human buyers. AIsa wants those interactions to become programmable, auditable, and settleable in fiat or stablecoins. Early traction is still young but notable: 50,000+ registered agents, 150x registered-agent growth from February to June 2026, and 200x API-call and transaction growth over the same period. Quick facts👇 ● founders: Jordan Liu ● total capital raised: $6.5M ● HQ: San Francisco, California ● Investors: Alibaba Group; Tribe Capital; Draper Associates If autonomous agents become buyers of digital resources, checkout may need to become agent-native too.

Savi Security just raised $7 million in seed funding to make scam protection real time for families. Acrew Capital led the round, with Magnify Ventures, TTCER Partners, and Resolute Ventures participating. The Los Angeles company is launching an iOS and Android app that screens texts, voicemails, unknown calls, live conversations, emails, and images before a user trusts the interaction. This is consumer cybersecurity moving closer to the moment of contact. Savi Security is not selling another awareness tool or after-the-fact report. It uses behavioral AI to detect scam patterns while the attack is unfolding, including an On Call feature that can quietly listen during a suspicious conversation and warn the user in real time. The founder story makes the wedge unusually concrete: Patrick Coughlin and Ryan Coughlin started the company after their mother was targeted by an AI voice-clone kidnapping scam. Scamwise, their free checker, has already reviewed nearly 100,000 suspicious submissions. Quick facts👇 ● founders: Patrick Coughlin; Ryan Coughlin ● total capital raised: $7M ● HQ: Los Angeles, California ● Investors: Acrew Capital; Magnify Ventures; TTCER Partners; Resolute Ventures As AI makes fraud cheaper and more personal, consumer security may shift from "teach people to spot scams" to "intercept the scam before trust breaks."

Kaon (prev. FlowGPT) just raised a reported $60 million Series B and is hiring to build AI-native entertainment around personalized story worlds. The San Mateo company closed the round with backing from B Capital, Redpoint Ace, Goodwater Capital, and DCM Ventures. The official announcement kept the amount undisclosed, while Variety and other coverage reported it at around $60 million. The interesting part is not just generated media. Kaon (prev. FlowGPT) is building a customization engine that writes, illustrates, and evolves story worlds around each user's intent, emotion, and behavior. Its flagship app Emochi reports more than 2 million daily active users spending an average of 150 minutes per day, with about $45 million in ARR. That makes this a consumer attention bet rather than another workplace automation round. If Kaon (prev. FlowGPT) can connect owned inference infrastructure, live user feedback, and entertainment loops, the product becomes less like a content catalog and more like a private story engine for every user. Quick facts👇 ● founders: Jay Dang; Alex Xi; Henry W. ● total capital raised: $70M+ reported ● HQ: San Mateo, California ● Investors: B Capital; Redpoint Ace; Goodwater Capital; DCM Ventures ● Partners: Nebius; DigitalOcean The category test is retention: AI entertainment will win only if personalization turns one story into many private worlds people keep returning to.

Most AI agents still run into a simple bottleneck: live web data is messy, rate-limited, and constantly changing. Oxylabs has raised $130 million from Warburg Pincus LLC at a $3.6 billion valuation to scale the infrastructure layer behind real-time web intelligence. What stands out is not only size. Oxylabs says it has reached $350 million in ARR, serves more than 350,000 tech teams, and runs infrastructure used by Fortune 500 and AI companies. Its products sit below the visible agent interface: proxies, scraping APIs, headless browsing, web index work, and compliance tooling that help systems access public web information reliably. Why now: agentic search turns fresh web access from a data-collection workflow into production infrastructure. Research agents, monitoring systems, pricing tools, fraud teams, and enterprise copilots all need current online context, not static indexes from yesterday. Quick facts👇 ● founders: Julius Černiauskas ● total capital raised: at least $130M ● HQ: Vilnius, Lithuania ● Investors: Warburg Pincus LLC The real shift: as AI moves from answering questions to acting on live information, the boring layer of compliant web data access may become one of the more valuable parts of the stack.

Augmodo's Smartbadges are starting to look less like retail hardware and more like an operating layer for physical work. The Seattle startup raised $21M in new funding led by existing investor TQ Ventures, valuing the company at $350M. The round extends a 2025 $37.5M raise and comes as Augmodo pushes its wearable spatial AI beyond store shelves into warehouses, factories, automotive settings, hospitals, facility maintenance, delivery operations, and training. The product edge is specific: lightweight dual-camera Smartbadges worn by workers passively build 3D maps and digital Realograms of physical spaces. In retail, that means live shelf inventory. In broader operations, it starts to become a data layer for tasks that still happen away from screens. The traction signal is why this financing stands out. Augmodo says revenue grew 10x over the past year, it maps more than 186M square feet of retail space monthly, and it expects to pass 1B square feet per month by year-end. It is also adding 50 to 100 store locations each month. Quick facts👇 ● founders: Ross Finman ● total capital raised: at least $63.8M ● HQ: Seattle, Washington ● Investors: TQ Ventures; Lerer Hippeau; Jefferson River Capital; Arena Holdings; Chemist Warehouse; New Fare; Interlace Ventures; Webb Investment Network The bigger question is whether physical AI wins first by augmenting workers already on the floor, not by replacing them with robots.

Verve Motion just raised undisclosed strategic funding and is hiring to bring powered exosuits into aviation ground operations. IAGi invested in the Cambridge company through a round joined by HexArmor, bringing Verve Motion's total capital raised to more than $55 million. International Airlines Group (IAG) will trial an aviation-specific version of the company's SafeLift exosuit, initially focused on baggage handlers. The product is concrete robotics for a very physical bottleneck. Baggage, cargo, warehouse, and distribution workers repeat high-strain lifts for entire shifts; Verve Motion's soft powered exosuit is designed to offload up to 40% of lifting strain while keeping the worker mobile. What makes the round interesting is the buyer signal. This is not a lab demo looking for a use case. International Airlines Group (IAG) brings live ground-handling environments, while HexArmor brings PPE distribution reach into industrial safety. If exosuits move from pilot programs into safety budgets, worker augmentation becomes part of operating infrastructure. Quick facts👇 ● founders: Ignacio Galiana ● total capital raised: more than $55M ● HQ: Cambridge, Massachusetts ● Investors: IAGi; HexArmor ● Partners: International Airlines Group (IAG) The category test is adoption: can wearable robotics become as normal for heavy lifting as gloves, helmets, and other PPE are today?

Hakimo is turning existing cameras into an always-on AI security layer. The Menlo Park company has closed a $12 million growth round led by Zigg Capital, with Neotribe Ventures, Vertex Ventures SE Asia & India, defy.vc, and Rocketship.vc participating. Total funding is now $32 million. Physical security is still full of passive video, false alarms, and guards asked to watch too many feeds. Hakimo’s bet is that computer vision can make existing cameras useful in real time, without a rip-and-replace hardware cycle. That matters because the traction is no longer just conceptual. Hakimo says revenue has tripled year over year for the second consecutive year, the customer base is above 300, headcount doubled, and customers have reported up to 60% fewer security incidents. The newer AI-Powered Forensic Search product also moves the platform beyond live monitoring into natural-language search across recorded footage. Quick facts👇 ● founders: Samuel Joseph; Sagar Honnungar ● total capital raised: $32M ● HQ: Menlo Park, California ● Investors: Zigg Capital; Neotribe Ventures; Vertex Ventures SE Asia & India; defy.vc; Rocketship.vc In physical security, the winner may be software that watches the cameras and turns noise into action.

This is a very clever solution for catching rocket boosters

Marker just raised $13 million in seed funding to build an AI-native word processor for people who still care about craft. Index Ventures led the round, with verified institutional participation from LocalGlobe, betaworks, Radical Ventures, and Tiny Supercomputer Investment Company. The product idea is sharper than another autocomplete layer. Marker is trying to make AI live in the margins of the writing process: a super-thesaurus, fact-checker, conversational editor, ideation partner, and revision aid that helps writers think without automating the whole document away. That matters because the word processor has barely changed at the same time AI has made generic text almost free. Jon Steinback and Ryan Bowman are taking the opposite bet: writers will pay for tools that protect judgment, voice, and revision rather than flatten them. Quick facts👇 ● founders: Jon Steinback; Ryan Bowman ● total capital raised: $13 million ● HQ: London, England ● Investors: Index Ventures; LocalGlobe; betaworks; Radical Ventures; Tiny Supercomputer Investment Company The contrarian angle: the next writing platform may win less by writing more for users, and more by helping them stay meaningfully in control of what they mean.

Tangos just raised $20 million in seed funding to make financial-crime investigations less manual. Red Dot Capital Partners led the round, with Leaders Fund, Clarim Ventures, VentureIsrael, SignalFire, Clutch Capital, Selah Ventures, and Bright Data joining. The company is building autonomous AI agents for AML, sanctions, fraud, beneficial ownership, and hidden-relationship investigations. The wedge is not more alerting; it is turning piles of evidence into examiner-ready case files with audit trails. That matters because financial institutions have spent years improving detection, but the investigative layer still depends on scarce specialists moving between tabs, documents, entity networks, and hypotheses. Tangos is betting that AI can own more of that middle work while leaving final accountability with investigators. Quick facts👇 ● founders: Eyal (Al) Azoulay; Eddie A. ● total capital raised: $20 million ● HQ: Tel Aviv, Israel ● Investors: Red Dot Capital Partners; Leaders Fund; Clarim Ventures; VentureIsrael; SignalFire; Clutch Capital; Selah Ventures; Bright Data The contrarian part: financial-crime AI may be less about replacing compliance teams and more about making every investigator operate with a much larger evidence bench.

Gradium has extended its seed round to $100 million with NVIDIA joining the cap table, seven months after the Paris voice AI startup came out of stealth. The company is building real-time voice infrastructure for AI agents and interactive products, with models and APIs for speech-to-text, text-to-speech, live translation, voice cloning, and on-device speech. The important part is latency. Voice agents do not fail only because they sound robotic; they fail because pauses, interruptions, misread codes, and slow turn-taking break the interaction. Gradium is trying to move voice from a demo layer into production infrastructure, with developers and enterprises as the target buyers. That makes the NVIDIA signal interesting. The funding is not just about another voice model. It is a bet that voice becomes a default interface for agentic software, and that some of the hardest work will sit closer to model serving, edge deployment, and real-time orchestration. Quick facts👇 ● founders: Neil Zeghidour; Olivier Teboul; Laurent Mazaré; Alexandre Défossez ● total capital raised: $100 million ● HQ: Paris, France ● Investors: NVIDIA; FirstMark; Eurazeo; DST Global; Eric Schmidt; Xavier Niel The next voice AI race may be decided less by who sounds best in a clip, and more by who can make real conversations feel instant, interruptible, and reliable in production.

Polysense just raised $10.7 million and is hiring. Food factories still run too much quality control on samples, manual checks, and late corrections. Polysense is taking a more physical AI route: cameras inspect every product on the line, models detect drift or foreign material in real time, and AutoControl adjusts equipment before waste piles up. Felix Capital led the oversubscribed seed round, with Fortino Ventures, Syndicate One, and 100IN participating. The capital follows a $2.2 million round in 2025 and a year of expansion from pilots into live deployments with major food manufacturers across Europe, the US, and the Middle East. The traction is interesting because Polysense is not just selling dashboards. Its system links perception to process control on existing lines. One potato processor cut peeling time by 45%, while bakery and packaging deployments show the same broader pattern: less variability, fewer manual interventions, better yield. Quick facts👇 ● founders: Yarne De Munck; Lucas Van Dijck; Jarne Bogaert ● total capital raised: $12.9M ● HQ: Ghent, Belgium ● Investors: Felix Capital; Fortino Ventures; Syndicate One; 100IN ● Partners: Agristo; Darta; Poppies Bakeries; Fourneo Flatbreads Food manufacturing may become one of the clearest tests for physical AI: can software close the loop fast enough to change what happens on the line, not just explain it afterward?