.webp)
Startup Intelligence Database
Discover what the world's best startups are building before it's announced publicly.
Data Extraction
Computer Vision
Model Evaluation
Prediction
Optimization
Scout Out
AI project management for residential contractors with automated takeoffs and proposals.
Residential contractors estimate jobs using spreadsheets and manual blueprint takeoffs. Scout Out automates takeoffs, cost estimation, and branded proposal generation in a single platform. The founder is a fourth-generation contractor with a UC Berkeley triple major and Amazon engineering experience.
RAG
Data Extraction
Computer Vision
Speech AI
Model Evaluation
RunAnywhere
Unified SDK for deploying AI models on mobile, web, and edge with local-first privacy.
On-device AI inference is fragmented across Apple, Google, and Meta SDKs. RunAnywhere ships one unified SDK with a proprietary MetalRT engine hitting 550 tokens/sec on Apple Silicon, plus hybrid local/cloud routing. Open-source core builds developer trust; managed cloud captures revenue.
RAG
Data Extraction
Computer Vision
Speech AI
Model Evaluation
Resonate
AI-native messaging with voice emotion recognition and agentic workflows in team chat.
WhatsApp and Slack bolt AI onto legacy architectures. Resonate is built AI-native from the ground up: voice messages with emotion extraction, live transcription, agentic workflows inside the chat, and generative video. The founder scaled infrastructure at Brex. Whether the market wants a new messaging app is the risk.
RAG
Data Extraction
Computer Vision
Model Evaluation
Optimization
Rhizome AI
Hallucination-free, citation-backed FDA regulatory intelligence for pharma and biotech.
Pharma regulatory teams spend weeks searching FDA databases manually. Rhizome reads up to 1,000 documents per question across FDA, EMA, MHRA, and PMDA, with word-level citations and zero hallucination tolerance. 2.5TB of regulatory data indexed, serving 18 organizations. Built by EvolutionaryScale's sole founding product engineer.
Data Extraction
Speech AI
Model Evaluation
Optimization
Fraud and Risk Scoring
Ressl AI
Autonomous AI agents automating Salesforce admin and CRM configuration for enterprises.
Salesforce admin is expensive, slow, and repetitive. Ressl deploys AI agents that handle CRM configuration, lifecycle automation, and operational workflows, with customers live in under a week. They also open-sourced an MCP security scanner, which is a smart trust-building move for the security-conscious Salesforce buyer.
Data Extraction
Speech AI
Model Evaluation
Prediction
Optimization
Reframe
AI agents automating end-to-end hardware procurement from BOM to vendor coordination.
Hardware procurement across overseas manufacturers involves tracking every part from raw material to assembly line across emails, group chats, and spreadsheets. Reframe's agents take a BOM and autonomously handle all vendor communication, coordination, and logistics through Slack.
Data Extraction
Speech AI
Model Evaluation
Prediction
Optimization
Revion
Voice-first AI automating technician admin in automotive dealerships to recover billable hours.
New model capabilities have made automotive operations newly automatable. Revion is close to the workflow where buyers already spend.
RAG
Data Extraction
Computer Vision
Model Evaluation
Prediction
REV1
Auto-generates manufacturing-ready 2D drawings from 3D CAD models using AI.
Mechanical engineers spend hours creating 2D drawings from 3D CAD models, following 300+ pages of GD&T tolerancing rules. REV1 generates manufacturing-ready, standards-compliant drawings from a single prompt. The founder was a mechanical engineer at Apple and Tesla and lived this pain daily.
Computer Vision
Prediction
Optimization
Simulation
Anomaly Detection
Remy AI
AI-powered bi-manual robots learning warehouse tasks from VR demos in hours, not months.
Warehouse picking robots cost too much for SMB 3PLs. Remy deploys bi-manual mobile robots at 50% lower cost by replacing mechanical complexity with intelligence, fine-tuning to each customer's environment from a few photos. The co-founder has an ML PhD from Oxford in computer vision and robotics.
Computer Vision
Speech AI
Model Evaluation
Prediction
Optimization
Return Signals
Proactively engages e-commerce customers post-purchase via SMS to prevent returns.
E-commerce returns cost brands 20-30% of revenue. Return Signals proactively texts high-value customers after delivery, resolves issues before they become returns, and converts refunds into exchanges. 61% engagement rate and 74% response rate on outbound SMS. The founder built production conversational AI in healthcare.
Data Extraction
Speech AI
Model Evaluation
Prediction
Optimization
Proximitty
Autonomous loan servicing with AI agents for collections, compliance, and borrower engagement.
Banks sit on billions in delinquent business loans and manage collections manually. Proximitty's AI agents automate the full servicing loop: collections calls, compliance monitoring, skip tracing, and borrower engagement. Reportedly onboarded 5 bank/fintech customers with $1B+ in delinquent loans within weeks of launch.
Data Extraction
Computer Vision
Speech AI
Anomaly Detection
Reinforcement Learning
RamAIn
Autonomous desktop agents automating any application using computer vision, fully on-device.
RPA tools break when the UI changes. RamAIn's agents use computer vision to understand the GUI itself, making them resilient to updates and capable of automating any application, including legacy software with no API. Everything runs on-device for privacy. Cross-platform: Windows, macOS Intel, and Apple Silicon.
RAG
Data Extraction
Speech AI
Model Evaluation
Prediction
Pollinate
AI agents that automate supply chain execution and procurement for manufacturers.
Supply chain teams waste 40% of their time on manual ERP data entry. Pollinate connects to the ERP, enriches the data, and generates custom agents for each workflow, starting with three-way match (invoice, PO, receipt) for procurement.
Data Extraction
Model Evaluation
Prediction
Optimization
Personalization
Polymorph
Delivers individualized engagement at the right time and channel using living user profiles.
Braze and Iterable personalize by cohort. Polymorph personalizes per user, treating every individual as a segment of one using continuously learning ML models. The difference is the same gap between batch and real-time processing.
Model Evaluation
Simulation
Synthetic Data
Reinforcement Learning
Workflow Agent
Polymath
Builds simulated worlds for training AI agents on complex digital workflows.
AI agent benchmarks like SWE-bench test narrow coding tasks. Polymath builds full simulated companies where agents must plan, code, test, deploy, and monitor software across real tools (Slack, GitHub, Linear). The training environment matches the complexity of the actual job.