.webp)
Startup Intelligence Database
Discover what the world's best startups are building before it's announced publicly.
Data Extraction
Computer Vision
Speech AI
Model Evaluation
Prediction
Shortkit
SDK that gives any app a TikTok-style video feed with adaptive streaming and AI captioning.
Glassdoor, DoorDash, and CNN all launched their own short-form video feeds because TikTok proved the format works. Shortkit gives any app the same capability via SDK: adaptive streaming, AI captioning in 50+ languages, auto moderation, and analytics. Built by a former YouTube Shorts infrastructure engineer.
Data Extraction
Computer Vision
Model Evaluation
Optimization
Simulation
Servo7
Wheeled humanoids learning industrial tasks from VR demos in hours with no coding.
Traditional industrial robots take months to program. Servo7's wheeled humanoids learn tasks from VR demonstrations in hours with no coding. The founders have military autonomous systems experience and demonstrated a working robot arm with a custom 55M parameter vision-action model at YC.
RAG
Speech AI
Model Evaluation
Optimization
Workflow Agent
Sila
Agentic workspace messaging embedding custom AI agents in team channels for collaboration.
Slack AI and Teams Copilot bolt assistants onto legacy messaging. Sila builds agents into the messaging layer natively, with persistent context, tool-use, and channel presence that feel like team members, not chatbots. $20.7M raised with SOC 2 in progress.
Data Extraction
Model Evaluation
Prediction
Optimization
Personalization
Sequence Markets
Institutional crypto execution aggregating 20+ exchanges with 3.3 microsecond latency.
Institutional crypto traders route orders across 20+ exchanges manually. Sequence aggregates them into one execution layer with 3.3 microsecond internal latency. The founders built execution systems at a $13B fund and the Toronto Stock Exchange. The play is becoming the FIX protocol for digital assets.
RAG
Data Extraction
Computer Vision
Model Evaluation
Optimization
Sitefire
Automates content so brands get cited by AI agents like ChatGPT and Perplexity.
SEO optimized for Google. GEO (Generative Engine Optimization) optimizes for ChatGPT, Gemini, and Perplexity. Sitefire runs hundreds of probing prompts daily to map which sources AI agents cite, then generates content and outreach to get brands into those citations. Already working with BMW.
Model Evaluation
Prediction
Optimization
Personalization
Anomaly Detection
Sentrial
Production monitoring for AI agents with time-travel debugging and ML-powered anomaly detection.
LangSmith tracks LLM calls. Sentrial tracks agent behavior: did the agent actually do what it said it did? Time-travel debugging, branching, and an Experiments feature for testing prompt changes on live traffic. Drop-in LangChain integration.
Computer Vision
Model Evaluation
Prediction
Optimization
Simulation
Seeing Systems
Modular strike drones with AI swarm control for single-operator autonomous missions.
Fiber-optic guided drones immune to electronic warfare jamming, paired with swarm software that lets one operator command multiple platforms. Field tested with UK Royal Marine Commandos, Estonian Military, and European Armed Forces. Built by a Jane Street engineer and a Cambridge roboticist.
Data Extraction
Computer Vision
Model Evaluation
Prediction
Optimization
RoboDock
Autonomous robots charging, inspecting, and managing electric vehicle fleet depots.
Electric vehicle fleet depots need charging, inspection, and fleet management 24/7. RoboDock's robots handle all three autonomously, plugging in and unplugging vehicles with vision-guided manipulation. $99/month RaaS pricing. Featured by Forbes as one of the most promising YC W26 startups.
Data Extraction
Prediction
Optimization
Fraud and Risk Scoring
Workflow Agent
Ruma Care
Automates prior auth for biologics, cutting denial rates by 50% for specialty clinics.
Biologic medications change lives for autoimmune patients, but prior authorization denials block access for months. Ruma Care cuts denial rates by 50% and staff time by 20x by automating the full PA cycle. The founders met in Yale's CS50 class and combine clinical empathy with engineering speed.
Data Extraction
Model Evaluation
Workflow Agent
Natural Language Processing
Ritivel
Automates regulatory medical writing with word-level traceability and reproducible outputs.
Regulatory medical writing (CTDs, CSRs, INDs) takes months and costs pharma companies millions. Ritivel drafts them in minutes with word-level traceability and 100% reproducible outputs, running entirely on-premise with zero cloud storage. The founders built AI copilots at Microsoft Research.
Data Extraction
Speech AI
Prediction
Personalization
Anomaly Detection
Robby
Equips home service techs with AI talking points and upsell insights via ServiceTitan.
Three Harvard Business School dropouts who spend their days riding along with HVAC technicians and studying for the EPA 608 certification exam. Robby equips techs with AI talking points and upsell insights via ServiceTitan. First day of launch: $40K in new business closed within an hour.
RAG
Computer Vision
Model Evaluation
Optimization
Simulation
Rubric AI
Post-training lab building rubric-based reward models and agentic frameworks for LLM alignment.
Post-training alignment (RLHF, reward modeling) is where model behavior gets shaped, but the tooling is fragmented. Rubric builds rubric-based reward models and ships open-source agentic frameworks used by teams at Cal.com and Trigger.dev. The founder was Staff Payments Lead at Meta across Instagram, WhatsApp, and Facebook.
RAG
Data Extraction
Model Evaluation
Workflow Agent
Natural Language Processing
Salus
Real-time guardrails that validate AI agent actions before execution to prevent unsafe outputs.
AI agents act before humans review. Salus validates every action before execution using evidence-grounded evaluators, reducing misalignment by 52% across 12 frontier models and recovering 58% of blocked actions with guided retries. Install via pip, configure in YAML.
Data Extraction
Computer Vision
Speech AI
Model Evaluation
Optimization
Samora AI
Multilingual AI voice agents in 12+ languages with 80% first-contact resolution.
Most voice AI companies target English-first markets. Samora deploys across 12+ languages with 80% first-contact resolution, serving a UN agency, financial services firms, and recruiters. The founder is a visually impaired Stanford CS grad who challenged India's education board to let blind students study science.
Data Extraction
Prediction
Optimization
Personalization
Fraud and Risk Scoring
Scheduling Wizard
Automates physician scheduling for academic medical centers with proprietary optimization.
Physician scheduling at academic medical centers is a constraint optimization nightmare that legacy tools like QGenda handle poorly. Scheduling Wizard built a proprietary optimization language for it and already has 19 departments across 14 hospitals including Mass General, Johns Hopkins, and UT Southwestern.