Agentic Systems Engineer
Design and build the multi agent systems behind client products and our own platforms: orchestration, tool layers, memory, evals, and the runtime that keeps agents inside their lane.
- TypeScript
- Python
- Claude API
- MCP
- vLLM
- Postgres
This is the core build seat for agentic work. Where the forward deployed roles carry systems into a client environment, this role designs the systems themselves: how agents are decomposed, how they call tools, how they hand off, how they remember, and how you know when they are wrong.
You will work across the Claude and OpenAI APIs, open weight models, and MCP, and you will spend a large share of your time on the parts that are not the model: the orchestration runtime, the eval harness, the cost and latency budget, and the failure handling that turns a demo into a product.
Design agent architectures
- Decompose problems into single agent, multi agent, or plain workflow designs based on the task, and defend the choice with measurements rather than fashion
- Build tool layers with typed schemas, bounded side effects, and permission models enforced in code, and expose them over MCP where multiple agents or clients consume them
- Design memory and context management: retrieval, summarization, prompt caching, and context budgets that hold up at real conversation lengths
Build the runtime
- Implement orchestration in TypeScript or Python: planning loops, handoffs, parallel tool execution, cancellation, and durable state across long running tasks
- Engineer for cost and latency with model routing between frontier and small models, caching, streaming, and batching, with per task budgets that alarm when exceeded
- Handle the failure modes: malformed tool calls, loops, refusals, stale context, and the model doing something technically valid and completely wrong
Prove it works
- Build eval suites from real traces, run them on every prompt, model, or tool change, and block regressions in CI
- Instrument traces so a product owner can read what an agent did and why, without an engineer translating
- Write up findings and reference designs that the rest of the team and our clients can build on
Required
- Five or more years of production software engineering in TypeScript or Python, with at least one agentic or LLM tool calling system you built and operated for real users
- Deep familiarity with at least one frontier model API (Claude, OpenAI, or Gemini), including tool use, structured output, streaming, and caching behavior
- You have built an eval harness and used it to make decisions, not just to produce a chart
- Systems instincts: you think about idempotency, timeouts, retries, and state before you think about prompts
- You can explain what an agent is and is not good at to a client without overselling either side
Helpful, not required
- Experience with MCP servers or clients, the Claude Agent SDK, or comparable orchestration frameworks
- Running open weight models locally or self hosted with vLLM or MLX, and knowing how tool calling degrades on smaller models
- Background in distributed systems, workflow engines, or durable execution
Also hiring.
Engineering Lead and Client Partner
Remote Delivery
A player coach seat that combines technical lead, delivery owner, and second seat on sales.
Forward Deployed Engineer, Agentic Commerce
Remote Applied AI
Embed with commerce clients and ship agent driven discovery, catalog, and checkout systems into production on their stack, not ours.
Forward Deployed Engineer, Enterprise AI Integrations
Remote Applied AI
Wire LLM agents into the ERPs, CRMs, and ticketing systems enterprises already run, inside the client environment, and make the integrations survive production.
Show us something you built.
Send a resume or profile with a short note about work you have shipped. Links to code, systems, or launched projects tell us more than a cover letter ever will.
careers@contracollective.com