A deeply hands-on engineering lead (~80% coding, ~20% leading) who owns the backend services, REST APIs, and AI/LLM systems behind Firmable's B2B data products, guiding two pods through their team leads. Core stack: Python/FastAPI, PostgreSQL, Snowflake, DynamoDB, and production LLM agents with evals, harnesses, and observability.
As Software Engineering Lead, you'll own the function that designs, builds, and operates the backend services, RESTful APIs, and AI/LLM systems powering every Firmable product. You'll lead two engineering pods through their team leads while staying deeply hands-on yourself. Reporting to the Head of Data & Services. This is not a people-management seat. ~80% hands-on architecture and engineering, ~20% leading the function. You whiteboard the system, write the reference implementation, then raise the whole team to that bar. If you want to step away from the code, this isn't it.
Responsibilities:
- Backend and API architecture: Firmable's Python services (FastAPI, Pydantic, Logfire): the secure, well-documented backbone every product, internal agent, and external customer depends on.
- Data modelling schemas, access patterns, and integrity at scale across PostgreSQL, Snowflake, and DynamoDB; AI/LLM services and agents production systems that are fast, idempotent, observable, and cost-bounded; every LLM call a named, versioned tool with prompts, schemas, traces, and evals.
- Harness engineering the agent runtime the team builds on: tool-calling and MCP server layers, structured-output validation, bounded agentic loops, sandboxed execution, human-in-the-loop checkpoints, replayable traces.
- Eval harnesses scorers, labelled sets, trajectory and multi-turn evals, LLM-as-judge calibration, regression suites in CI, drift detection on vendor model updates.
- Scale and reliability systems serving millions of API calls: idempotency, caching, rate limits, cost ceilings, circuit breakers, bounded retries, graceful degradation.
- Observability: standard structured traces, p50/p95/p99 by stage, cost per request, and the dashboards that make it legible.
- Function leadership: lead and grow two pods through their leads; run architecture and code reviews; set the bar for quality, testing, and AI-native practice.
- Cross-functional glue: connect Data Platform, Application, and Product; turn early ideas into architecture decisions and working prototypes fast.
Requirements:
- 7-10 years of hands-on backend engineering, including time leading engineers or engineering leads, with 3+ years of hands-on experience with Python.
- Built and operated large-scale production systems with millions of API calls, and you know what breaks first when load grows.
- Expert in Python with deep production experience in FastAPI.
- Shipped AI/LLM services, APIs, and agents in production and stood up the evals that keep them trustworthy.
- Harness engineering experience you've built the scaffolding agents run inside: tool-calling frameworks, MCP servers, eval and test harnesses other engineers depend on, and you can walk us through it in detail.
- Hands-on data modelling across SQL and NoSQL.
- Designed and consumed RESTful services at scale.
- Shipped real work with agentic IDEs Claude Code, Cursor, or equivalent. Not "tried it" your default mode.
- Lead from the front: whiteboard to reference implementation to a team operating at that standard.
- Highly Valued Eval frameworks (Braintrust, Inspect AI, Promptfoo) and tracing stacks (Logfire, OpenTelemetry).
- Agent frameworks and protocols: Claude Agent SDK, Pydantic AI, LangGraph, MCP AWS and/or GCP at scale ECS/Fargate, Lambda, S3, RDS, DynamoDB B2B data, entity resolution, or high-throughput enrichment pipelines; Snowflake and modern data-platform tooling.
- Startup or scaleup experience where you shipped fast and owned outcomes end to end.