Founding AI Engineer
CodeRound is hiring for this role.
- Founding AI Engineer
- AI Agents
- LLM Agents
- Tool Calling
- MCP
- Agent Harness
- Browser Agents
- Computer Use
- Playwright
- Evals
- LLM Evaluation
- Langfuse
- OpenTelemetry
- Python
- asyncio
- FastAPI
- AWS
- ECS
- SQS
- Celery
- CI/CD
- Context Management
- Memory
- Structured Outputs
- Consumer AI
- Early Stage
- Gurgaon Onsite
Founding AI Engineer at Valo by Persin (tellvalo.com), Gurgaon, full time, 5 days in office, working directly with the founders (ex-COO Zomato; ex-Head GenAI IDFC Bank and Principal DS Zomato; ex-Head GenAI Kapture CX; product lead ex-Grab, Microsoft, Airtel). Valo is the personal assistant every Indian can hire: it takes and makes calls, messages, books and pays, and comes back done. The hire owns certainty of outcome: the agent harness (agents, tools, memory, a scheduler that keeps a task alive for days, gates for what Valo may do without asking, and evals) that takes a task from one sentence to a verified result in India, where calls drop, browser agents fail, portals want OTPs and payments must never run twice. Seed-funded, six people, building its own models on real task outcomes.
What you'll do
- Design the architecture that takes a task from a sentence to a verified result: agents, tools, memory, long-running scheduler.
- Own the gates that decide what Valo may do without asking, and never run a payment twice.
- Build browser and computer-use agents that finish real tasks through login walls, OTPs and dynamic pages.
- Build evaluation and observability: trajectory logging, replayable runs, LLM-as-judge scoring, regression gates in CI.
- Turn every finished task into a training example for Valo own models.
Must have
Built and ran production LLM agents at scale owning the full harness: multi-step tool calling and MCP servers, context and memory management, structured outputs, streaming. Built browser or computer-use agents (Playwright or CDP) that complete real tasks on third-party websites through login walls, OTPs and dynamic pages. Built evaluation and observability for agents: trajectory logging, replayable runs, LLM-as-judge scoring and regression gates in CI, tracing via Langfuse or OpenTelemetry. Strong Python (asyncio, FastAPI or Django) with production AWS (ECS, SQS or Celery workers, autoscaling, CloudWatch) and CI/CD. Willing to work 5 days in office in Gurgaon.
Good to have
Payments, fintech or consumer-scale agent work. Post-training or evaluation on task outcomes (SFT/RL). Strong college pedigree and early-startup or founding-engineer experience.