Agentic managed talent marketplace.
75K professionals across marketing, design, operations, and engineering. A matching engine trained on 7+ years of real placement data. Agent-first operations managing the full engagement — contracts, time tracking, invoicing, and payments.
Seven years of iteration.
Agentic Managed Marketplace
Pangea is an agentic managed talent marketplace — 75K professionals across marketing, design, operations, and engineering. We manage the full engagement: contracts, time tracking, invoicing, and payments. Not just matching — we run the relationship.
Matching That Learns
The matching engine is trained on 7+ years of real hiring outcomes — semantic search, ELO ranking, LLM reranking. Every interview happens in our platform. Every transcript is analyzed. Every client response feeds back in. It's a reinforcement learning system that gets measurably better with every placement.
Agent-First Operations
An agentic layer manages the pipeline end-to-end — from job creation to invoicing. The architecture ships new agents in days. Standard stack (Node.js, Postgres, React, GCP), portable, zero lock-in.
Unit economics that compound with automation.
Backed by Y Combinator, Pangea optimized for unit economics and platform leverage — the metrics that compound post-integration. 24x revenue-per-placement growth and margins expanding to 27%, all before the full agent layer is deployed. Production numbers, not projections.
Everything below is running in production today. Book a call with Adam — we'll open the product and the codebase.
Seven steps. One platform.
Every step from first form submission to final payment is built and running natively. AI agents layer on top to automate the highest-leverage workflows.
Meeting Agent
$0.13 / meetingReplaces ~2 hours of post-meeting ops per call. Extracts insights, updates the knowledge base, and files tickets — no human involvement. At scale, thousands of hours reclaimed annually.
Job Posting Agent
Minutes, not hoursA single sales call transcript becomes published job posts — company profile, requirements, and live listing. Scales job creation across every customer without adding headcount.
Match Agent
Hybrid search + LLMRuns BM25 + semantic retrieval across 75K profiles, applies performance ranking, then LLM-reranks to a 3–5 candidate shortlist — end to end in minutes.
Roadmap
ShippingJob Manager · Contract Manager · Sales Qualification — same proven pattern, new workflows.
75,000 profiles to 3–5 top candidates.
Hybrid search (BM25 + semantic) combined with performance ranking and LLM reranking. Three independent layers produce a single ranked shortlist — portable across any talent vertical.
Qualification Funnel
Three people built this. That's the point.
Three engineers own the full stack — backend, matching engine, agent layer, and infrastructure. The founder transitions to an advisory role; the technical team is available for full integration. Zero organizational overhead, direct access to every decision that shaped the platform.
Brown University & MIT. Built Pangea from zero to a production platform serving enterprise clients including Bechtel and SecurityScorecard. Leads product, strategy, and go-to-market.
Applied Mathematics, Brown University. Architected the full technical stack — matching engine, agent infrastructure, and platform backend.
Leads creative direction and AI implementation across the platform, from product design to agent development.
Seven years of iteration. Ready to integrate.
Standard stack, no lock-in, no proprietary dependencies. Everything here is running in production today. We're flexible on structure — asset acquisition, technology license, or team integration.
Profiles with embeddings
Already indexed and ready for matching on day one.
Production agents running
Meeting Agent and Job Posting Agent. Reusable pattern for building more in days.
Years of platform iteration
Contracts, payments, time tracking, invoicing — all running in production for 7+ years.
Proprietary dependencies
Standard stack: Node.js, PostgreSQL, React, GCP. Deploy on any cloud.
The moat is in the data. Seven years of placement history power the ELO ranking system. 75K profiles have vector embeddings tuned to actual hiring outcomes. Two production agents process real transactions daily. The matching engine improves with every placement — that feedback loop and the data behind it are what take years to build.
Backend
- Node.js / TypeScript
- Koa.js
- PostgreSQL + TypeORM
- Redis + RabbitMQ
Search & AI
- Elasticsearch 7.17
- Pinecone vector search
- Claude Agent SDK
- DSPy + OpenAI embeddings
Frontend
- React 18 + Vite
- MUI component library
- Redux
Admin
- Next.js 15 (App Router)
- React 19 + shadcn/ui
- React Query
Infrastructure
- GCP Kubernetes
- Railway
- Docker
Payments
- Stripe billing
- Invoicing + payouts
Analytics
- PostHog
- Customer.io
- Segment
Knowledge
- Markdown / Git KB
- Auto-updated by agents
- Grows with every interaction