.webp)
Startup Intelligence Database
Discover what the world's best startups are building before it's announced publicly.
Model Evaluation
Simulation
Synthetic Data
Reinforcement Learning
Natural Language Processing
Traverse
Builds training data for AI in ambiguous domains like law, healthcare, and strategy.
Scale AI labels verifiable data: is this a cat or a dog? Traverse builds training environments for non-verifiable work: legal judgment, medical decisions, sales strategy. The moat is expert human reasoning data that cannot be synthetically generated. Co-founders also run Clice AI (YC W26).
Data Extraction
Computer Vision
Model Evaluation
Prediction
Optimization
Unifold
Universal API for multi-chain crypto deposits, abstracting away chains, gas, and compliance.
Crypto deposits require users to navigate chain selection, gas fees, and wallet complexity. Unifold abstracts it all into a single API that accepts deposits across Ethereum, Solana, Bitcoin, and more. The founder's previous company (Streambird) was acquired by MoonPay and became MoonPay Wallets.
Data Extraction
Computer Vision
Prediction
Optimization
Personalization
Terranox AI
AI-powered uranium exploration using physics-informed models on 70+ years of geology data.
Nuclear energy is back on the geopolitical agenda and uranium supply has not kept up. Terranox applies physics-informed ML to 70+ years of geological data to find deposits that human geologists missed. The CEO published planetary discovery methods adopted by NASA and spent 3.5 years at BCG on nuclear and mining strategy.
Data Extraction
Model Evaluation
Personalization
Recommendation Systems
Multimodal AI
Sonarly
Autonomous AI that triages and fixes production incidents with contextual pull requests.
Sentry detects errors. Sonarly detects, investigates, and fixes them autonomously with contextual PRs that reference the production state. The agent updates its understanding of the codebase after every incident, so it gets better at fixing your specific system over time.
Data Extraction
Model Evaluation
Prediction
Optimization
Personalization
Sponge
Financial infrastructure enabling AI agents to autonomously hold, manage, and spend money.
AI agents need to spend money. Sponge gives them programmable wallets with spending controls across fiat and crypto. The founding team built crypto payments at Stripe. First-class MCP support means Claude and other agents can transact through Sponge natively.
RAG
Computer Vision
Speech AI
Model Evaluation
Optimization
Stilta
Agentic AI automating prior art search, claim analysis, and drafting for patent attorneys.
Patent attorneys bill $500-1,000/hour for prior art search, claim analysis, and drafting. Stilta automates these tasks with multi-agent orchestration, already deployed at leading law firms in Europe and the US. Four McKinsey/QuantumBlack co-founders with elite AI engineering backgrounds.
RAG
Data Extraction
Computer Vision
Model Evaluation
Prediction
Synthetic Sciences
Claude Code for Science, automating code generation and research workflows.
Computational scientists spend more time writing data pipelines and analysis scripts than doing science. Synthetic Sciences plugs Claude Code into 30+ biomedical APIs (PubMed, ChEMBL, ClinicalTrials.gov) with modular Agent Skills for reproducible workflows. Claude Code for Science is a clear positioning.
Model Evaluation
Prediction
Optimization
Personalization
Fraud and Risk Scoring
SpotPay
Stablecoin wallet for instant, low-cost cross-border payments for expats and freelancers.
Cross-border payments cost 6-8% through traditional remittance providers. SpotPay routes through stablecoins with a Visa-compatible card (Apple Pay, Google Pay) in 40+ countries, growing 43% WoW on TPV. The founder was a tech lead at Google (YouTube Monetization) and grew up navigating broken Caribbean payment rails.
RAG
Data Extraction
Computer Vision
Model Evaluation
Prediction
Strand AI
Predicts missing biological data from routine samples to enrich sparse clinical trials.
Clinical trials generate sparse data because advanced assays (spatial transcriptomics, proteomics) are too expensive to run on every patient. Strand predicts missing modalities from routine samples, and their model beats state-of-the-art for spatial proteomics prediction. Built by researchers who created petabyte-scale spatial biology platforms.
RAG
Data Extraction
Speech AI
Model Evaluation
Prediction
Syntropy
Agentic coding from spec to tested PR for enterprise codebases with 10K+ lines.
AI coding agents write code well but struggle with complex features that span multiple services. Syntropy adds a collaborative spec layer where a human and an advisor agent plan together before code generation starts, targeting enterprise codebases with 10K+ lines. Stanford CS founders with Apple and Amazon backgrounds.
Data Extraction
Computer Vision
Model Evaluation
Prediction
Fraud and Risk Scoring
Sparkles
Lets non-technical team members safely contribute to codebases with AI-guided editing.
Marketers and designers constantly need small code changes (copy updates, config tweaks) but cannot touch the codebase. Sparkles gives them a sandbox connected to GitHub where AI guides them through edits and creates PRs. Built by a 20-year-old Ukrainian solo founder who dropped out of UCL.
Data Extraction
Computer Vision
Model Evaluation
Prediction
Personalization
Skillsync
Discovers elite engineering talent by analyzing real code contributions instead of resumes.
Resumes are self-reported and unreliable. Skillsync evaluates engineers by analyzing their actual GitHub contributions: code quality, collaboration patterns, domain expertise. The founders are elite open-source contributors who experienced the hiring mismatch firsthand.
Model Evaluation
Prediction
Optimization
Simulation
Personalization
Squid
AI-powered collaborative platform for electric grid planning and scenario simulation for utilities.
Electric grid planning runs on file-based models that do not version, share, or collaborate. Squid replaces them with a browser-based AI platform built by engineers from National Grid and Octopus Energy. Already deployed at both companies. SOC 2 Type II and ISO 27001 certified.
RAG
Data Extraction
Computer Vision
Speech AI
Model Evaluation
Shofo
World's largest structured video library and social media data API for AI model training.
AI labs need video training data but most of what exists online is unstructured and unlicensed. Shofo structures millions of hours of video and exposes social media data via API. The team previously built Correkt (multimodal search, 40K+ users), so the indexing infrastructure already exists.
Data Extraction
Model Evaluation
Fraud and Risk Scoring
Anomaly Detection
Natural Language Processing
SideKit
Privacy-first unified mobile SDK for analytics, version gating, and messaging in one line.
Firebase is free and deeply integrated, but it tracks everything. SideKit bundles analytics, version gating, and in-app messaging into one privacy-first SDK with zero user data tracking by default. Open-source (MIT), one-line integration. A lightweight bet for the post-ATT privacy era.