Forward Deployed AI Engineer at Puente Talent Partners

Job Description

Sit with the client's operators and execs, ask sharp questions, and turn fuzzy automation requests into concrete, shippable plans with clear judgment on what not to automate.
Build production AI agents with real error handling, retries, idempotency, fallbacks, and human-in-the-loop checkpoints across orchestration tools and custom code.
Build the intelligence layer using LangChain, LangGraph, RAG, structured outputs, and voice agents where they fit the use case.
Integrate with the real world: REST APIs, webhooks, OAuth2, CRMs, ERPs, Supabase, Airtable, and Google Workspace to execute write-access operations on live records.
Own production reliability: debug live agents, instrument logging, track costs, set failure alerts, and keep systems running while clients watch.
Hold the client relationship by explaining tradeoffs to non-technical stakeholders, pushing back on requests that do not make sense, and keeping them confident to keep buying.
Codify each build into reusable templates so the next deployment ships faster. Who You Are
You have 3+ years shipping production software, including real AI and agent work delivered to actual users or clients — you have built it, deployed it, and watched it run in the wild.
You have shipped at least one real project end to end, scoped it, built it, and owned it when it broke in production.
You work as an army of one: you scope, build, debug, and ship without hand-holding, formal specs, or a manager pointing the way.
Your English is excellent in speech and writing — you run calls with US founders every week and explain technical tradeoffs to non-technical stakeholders with clarity.
You think in systems and have wrestled production AI against brittle reality: auth failures, flaky data, rate limits, and the unglamorous 80% that is not the happy-path demo.
You have strong Python plus at least one of JavaScript or TypeScript, and you are fluent with OpenAI and Anthropic APIs, prompt engineering, structured outputs, RAG, and evals. Bonus Points
Prior work at an AI-automation agency or as a forward-deployed or solutions engineer.
Shipped Vapi or Retell voice agents in production.
Experience with vector databases like pgvector or Pinecone.
Docker and containerization.
Former technical founder or freelancer who carried your own clients. What Success Looks Like