multimodal-ai
(47)
Luma AI Wants You to Direct, Not Just Prompt
Luma has grown from a five-second video curiosity into a serious creative workbench. Ray3.2 gives filmmakers real control, while the credit meter gives everyone a reason to plan before pressing generate.
Hailuo AI Makes the First Cut. You Still Direct.
MiniMax’s pocket video studio can turn a prompt and a pile of references into a polished short with sound. The trick is knowing when to applaud the take, when to regenerate it, and when to call an editor.
Microsoft AI Bought a Startup’s Brain. Now It Has to Build Its Own.
Microsoft spent about $650 million to import a ready-made AI team, then gave it Copilot, Bing and an enormous distribution machine. Two years later, the wager is no longer about catching ChatGPT - it is about whether Microsoft can turn borrowed momentum into models, products and trust of its own.
OneReach.ai Wants to Be the Air-Traffic Control for Your AI Agents
Companies learned how to launch an AI agent. Now comes the expensive part: stopping a thousand of them from colliding. OneReach.ai is betting the durable business sits in orchestration, governance and the connective tissue between models, channels and old enterprise systems.
OpusClip Paid People to Use Its First Product. Then One Feature Found 5 Million Users
The livestreaming app stalled at roughly 200 users. Its clipping tool did the opposite - revealing a useful lesson about product-market fit, creator software and the limits of one-click editing.
Ananth Kashyap Is Teaching AI to Get Its Hands Dirty
The Sureform founder started with a high-school iOS app and arrived at a harder question: how do you teach AI what real work feels like? His answer begins in the wonderfully untidy places where people already know what to do.
Andriy Mulyar Is Teaching AI to Read the Blueprints
The Nomic founder went from clinical language models to open-source AI on a laptop. Now he is narrowing the map, building agents that can navigate the drawings, specifications and institutional memory behind the physical world.
Luyen Chou Is Teaching AI to Pay Attention
After nearly four decades moving between classrooms, school launches and product suites, the DeweyLearn co-founder is betting that useful AI begins with a modest act: watching how people actually learn.
DeweyLearn Put an AI Assessor in the Room - Now It Has to Earn the Expert's Trust
The New York startup moved beyond chatbot-style tutoring to judge what people actually do - from knife work to therapy practice. A $5 million Series A gives it room to prove that expert feedback can scale without becoming generic surveillance.
Brilliant Labs Keeps Rebuilding the Same $349 Bet: AI Glasses You Can Rewrite
The Singapore startup learned that putting AI in your sightline is easy to demo and hard to live with. Halo is its third attempt - thinner, longer-lasting, more private, and open enough for developers to make the glasses useful in ways the company has not imagined.
HyNote Built 20 Apps Before It Found the One Million-User Habit: Stop Taking Notes
The small Palo Alto team stopped treating transcription as the product. HyNote's wager is that the winner will own the whole trip from a whispered thought or messy PDF to an action item, quiz, slide deck or shareable page.
Ivan Poupyrev Is Teaching Machines to Read the World Beneath the Screen
For three decades, Ivan Poupyrev has kept asking the same practical question: how can computing escape the glass rectangle? At Archetype AI, his answer begins with the signals already humming through factories, streets, clothes, and machines.
Lava Kumar
Lava Kumar is the co-founder and Chief Product Officer of Entropik, a Bengaluru-based Emotion AI company that reads human reactions across facial expression, voice, gaze and attention. After nearly two decades building products at Yahoo, Motorola, Wipro, Operative and Integral Ad Science, he returned from the US in 2016 to co-found Entropik and turn affective computing into a working business tool. He leads Decode, the company's integrated human insights platform, and has become a recognizable voice arguing that market research should measure how people actually feel, not just what they say.
Bharat Singh Shekhawat
Bharat Singh Shekhawat is a co-founder and Head of Engineering at Entropik, the Bengaluru-based human insights company he helped start in 2016. He builds the systems that turn signals from faces, voices, and eye movement into measurable read-outs of how people feel and decide, and has spent more than fifteen years designing data-intensive platforms and turning research ideas into shipping products.
Shofo
Shofo is a San Francisco startup (Y Combinator W26) building what it calls the world's largest indexed video library - a 'Common Crawl for video.' It continuously crawls billions of public short-form clips from platforms like TikTok, Instagram Reels and YouTube Shorts, then runs them through a pipeline that sanitizes, deduplicates, segments and labels footage with object detection, activity recognition and reasoning annotations. Shofo sells custom, ready-to-train datasets to AI labs, delivering a request like '100K hours of cooking videos where someone is holding a pan' in days rather than months.
Pathos AI
Pathos AI is an AI-driven biotech company building what it describes as one of the largest multimodal foundation models in oncology. By combining millions of clinical, molecular, and imaging records with proprietary AI, Pathos aims to identify the right patients for a given therapy, design smarter adaptive clinical trials, and de-risk cancer drug development. Founded in 2020 by Eric Lefkofsky and Ryan Fukushima and led since 2025 by CEO Iker Huerga, the company raised a $365 million Series D that valued it at roughly $1.6 billion.
Eitan Cohen
Eitan Cohen is the CEO and co-founder of TechSee, a Tel Aviv-based company that pioneered visual and agentic AI for customer service. Trained in computer science and psychology at Tel Aviv University, Cohen has spent two decades building and selling technology companies. He started TechSee in 2015 after realizing that voice and text support could not solve the everyday frustration of guiding someone through a technical problem they could see but not describe. Under his leadership, TechSee's computer-vision and augmented-reality platform now serves more than 1,000 customers, including Fortune 500 brands, and reports avoiding millions of technician visits each year.
Kaliber Labs
Kaliber Labs is a San Francisco AI research foundry building software that captures rare human expertise - starting in the operating room - and turns it into scalable tools. Trained on one of the world's largest libraries of arthroscopic surgical video and guided by dozens of leading orthopedic surgeons, Kaliber's systems perceive anatomy in real time, automate surgical documentation, and give surgeons, staff, and patients AI that understands what is actually happening inside a procedure.
SightCall
SightCall is a San Francisco- and Paris-based enterprise software company that turns a customer's or technician's smartphone camera into a live support channel. Its cloud platform blends real-time WebRTC video, augmented-reality annotations, and multimodal generative AI so that a remote expert - or increasingly an AI agent - can see what someone in the field sees, guide them through a fix, and automatically document the session. Used by brands like Ford, Allianz, GE HealthCare, Deutsche Telekom and Lavazza, SightCall targets field service, insurance claims, customer service and healthcare, aiming to resolve problems on the first contact and cut unnecessary truck rolls.
Integral AI
Integral AI is a Japan-US startup building foundation world models - AI systems meant to learn, plan, and act in the real world the way people do, with an emphasis on energy efficiency and continuous, autonomous skill learning. Founded in 2021 by ex-Google AI leaders Jad Tarifi and Nima Asgharbeygi, the company positions its interactive world models as infrastructure for an AI-based economy, spanning robotics, universal simulators, and multimodal interfaces. In December 2025 it made headlines by claiming to have tested the world's first AGI-capable model, TOWA.
Meadow AI
Meadow AI is a Seattle-based multimodal AI platform that gives restaurants and retailers real-time visibility into their physical operations. By combining computer vision, natural language processing, and data from existing point-of-sale, labor, inventory, video, and audio systems, Meadow automates the manual 'secret shopper' audits that chains have long done by hand - continuously evaluating customer experiences and staff performance, then surfacing AI recommendations to lift revenue, tighten consistency, and reduce labor costs. The company emerged from stealth in November 2025 with $6M raised and more than $2.5M in contracted annual recurring revenue.
Neural Defend
Neural Defend is a New York and Delhi-based cybersecurity startup building multimodal deepfake detection for enterprises. Founded in 2024 by MIT- and Intel-affiliated researchers, its proprietary AI spots AI-generated faces, manipulated video, cloned voices, and forged documents in real time - delivered as developer-friendly APIs and SDKs, plus on-premises deployments for banks, fintechs, and government. The pitch: make catching a deepfake as easy as an API call.
Neuralzome Cybernetics
Neuralzome Cybernetics is a Bengaluru-based deeptech startup building teachable, no-code AI agents that let ordinary off-road vehicles drive themselves through messy, GPS-denied places like orchards, farms and job sites. Its flagship autonomy stack, NeuralPilot, fuses vision, SLAM and GPS with a layered safety system and learns from human demonstration, while its RedPill simulator trains robots on photorealistic digital twins before they ever touch dirt. Founded in 2023 by Mohan Sivam and Aditya Shriwastava, the company sells autonomy as a service to labor-intensive industries and raised $2.4M in pre-seed funding in August 2025 to expand into North America and Europe.
Roe AI
Roe AI builds agentic AI that investigates fraud, disputes, and anti-money-laundering (AML) cases end to end - reading documents, images, websites, and structured data to produce cited evidence and audit-ready dispositions. Founded in 2024 by ex-Snowflake and ex-Meta engineers and backed by Gradient Ventures and Y Combinator, the company started as an AI-powered data warehouse for querying unstructured data with SQL and sharpened into a financial-crime investigation platform used by teams at Affirm, eBay, and a Global Systemically Important Bank.
Chang She
Chang She is the co-founder and CEO of LanceDB, the company behind the open-source Lance columnar format and a database built for multimodal AI. A co-author of the pandas library alongside Wes McKinney, he spent nearly two decades building data tooling - from quant research at AQR to the DataPad startup (acquired by Cloudera) to VP of Engineering at Tubi - before setting out to replace the aging Parquet-and-Spark stack with what he calls the Multimodal Lakehouse. In June 2025 LanceDB raised a $30M Series A led by Theory Ventures, with adoption at Runway, Midjourney, and Character.ai.
Max Sim
Max Jai Sim is the co-founder and CEO of Meadow AI, a Seattle startup that uses multimodal AI - computer vision, natural language, and operational data - to give restaurants and retailers real-time visibility into their physical locations and to automate the manual 'secret shopper' audits chains still run by hand. A Harvard-trained applied physicist, he previously co-founded the home-closing startup Modus, acquired by Compass in 2020, and earlier worked at T-Mobile and the lunch-delivery startup Peach. Meadow emerged from stealth in late 2025 with $6M raised, including a $4.5M seed round, and more than $2.5M in contracted annual recurring revenue.
Mohan Sivam
Mohan Sivam is the founder and CEO of Neuralzome Cybernetics, a Bengaluru deeptech startup teaching robots to drive themselves across orchards, farms and job sites no GPS map can describe. His pitch is deceptively plain: robots should learn by watching a human do the job once, not by waiting on a team of PhDs to code every turn. Neuralzome's NeuralPilot autonomy stack, RedPill simulation engine and rugged Jeeno platform aim to put off-road robots to work where labor is scarce and the terrain is unforgiving. In August 2025 the company raised $2.4M in pre-seed funding led by 8X Ventures. A serial builder who previously co-founded Flo Mobility and ran GAMASOME, Sivam describes his mission with a phrase that sticks: making robots sentient.
Thoth AI
Thoth AI is a Singapore-headquartered global AI data solutions company that builds the high-quality datasets, evaluation frameworks, and human-in-the-loop operations that production AI systems depend on. Spanning 200+ languages and 100+ countries, it handles data collection and annotation across images, video, 3D point cloud, text, and audio, along with RLHF, model evaluation, content moderation, and multilingual customer experience for teams building LLMs, VLMs, multimodal systems, and applied AI like robotics.
Typeface
Typeface is an enterprise generative-AI platform that helps marketing teams plan, create, personalize, and launch on-brand content and campaigns at scale. Founded in 2022 by former Adobe CTO Abhay Parasnis, it grounds AI in a company's brand guidelines, layouts, and audience data so output stays consistent and safe. Its flagship product, Typeface Arc, combines specialized marketing agents, a brand-aware knowledge graph, and collaborative workspaces to run the full content lifecycle. The company reached a $1 billion valuation in 2023 and counts brands like Asics, Johnson Controls, Sally Beauty, and Microsoft among its users.
Spiral
Spiral is a New York-based data infrastructure company building a database reimagined from the ground up for the AI era - what it calls the 'Third Age of Data,' when machines, not humans, are the primary consumers of data. Built on Vortex, an open-source columnar file format donated to the Linux Foundation, Spiral streams data directly from object storage into GPU memory, unifies governance across every data type from tiny embeddings to massive video, and aims to keep expensive GPUs saturated instead of idle. Founded by ex-Palantir and ex-Citadel engineers, it emerged from stealth in September 2025 with $22M in Seed and Series A funding led by General Catalyst and Amplify Partners.