What we build: An agentic middleware platform: multi-agent orchestration, async RAG pipelines for unstructured data, just-in-time context retrieval via MCP tools, sandboxed code execution, and token-optimized tool calls with progressive discovery.Your mission:- Eval pipeline: Improve our eval pipeline for complex multi-step agentic systems- Agentic core components: Work on memory and cross-project memory architectures- Document layout processing: Advance processing for complex, unstructured inputs- R&D with production impact: Research-heavy work with the opportunity to push into production codeStack: Python 3.13 async, FastAPI, PostgreSQL, DynamoDB, Turbopuffer, AWS, LLMs (multi-provider), in-house durable execution engine.What we're looking for: Solid ML fundamentals; strong Python (clean code beyond notebooks); research ability (read papers, search literature, implement fast); curiosity; highly appreciated: experience with LLMs/agents, async Python, AWS, or open-source contributions.Why this internship: Open problems (eval, memory, document understanding); direct collaboration with ML Engineer Lead and CTO; modern stack and problems 2 years ahead; small team with ownership.Interview process: Apply (hello@cominty.ai) → Intro call (with our CTO) → Technical assignment → Final chat (with co-founder).What Cominty offers: High-impact role; direct founder collaboration; rapid career growth; modern work environment (2–4 days remote/week); attractive compensation.