ABOUT THE ROLE
Tight Line LLC is a software consultancy that takes on hard problems for enterprise clients. We're looking for a contract AI engineer to work on short-term client projects, typically a few weeks to a few months each, across multiple enterprise engagements. The work is Python-first for agent backends, with JavaScript/TypeScript for UIs and the occasional all-TS build: building AI agents and agentic workflows that run in production, integrating with real CRMs, ERPs, databases, and legacy systems, not proof-of-concept demos.
You'll often be embedded directly with client teams, so this is as much a consulting role as an engineering role. You should be comfortable joining a project mid-stream, earning trust quickly, and leaving the client's team better than you found it.
WHAT YOU'LL DO
* Design and build agentic workflows using LangChain, LangGraph, DeepAgents, and LangSmith
* Build MCP servers and register them with agents (the MCP client side) so tools and data sources plug in cleanly
* Build and harden RAG pipelines end to end: ETL and ingestion, chunking, embedding, vector stores, hybrid retrieval, and evaluation
* Instrument and evaluate agent behavior (tracing, evals, regression suites) so "it works" is something you can prove, not just claim
* Integrate agents with client systems and data sources, including legacy ones
* Pair with client engineers, explain your decisions, and transfer knowledge so clients can own the system after you roll off
* Participate in client meetings: demos, status updates, technical discussions with both engineers and stakeholders
WHAT WE'RE LOOKING FOR
* Strong Python engineering skills plus solid JavaScript/TypeScript. The LangChain SDK supports both; Python tends to lead for agents and JS/TS for UIs. Either way, you'll be writing production code: typed, tested, and packaged for someone else's team to maintain
* Solid CS fundamentals: data structures and algorithms, and a real grasp of concurrency with async/await in both Python and JS/TS
* Comfortable building HTTP clients against whatever a client system speaks: REST, JSON-RPC, gRPC, GraphQL
* Testing discipline: TDD where it fits, and at minimum fluent with unit, integration, and system tests, including mocking and stubbing
* Demonstrated, shippable work building AI agentic workflows. You can show us something you built and talk in depth about the design decisions, failure modes, and how you evaluated it
* Strong hands-on experience with the LangChain ecosystem: LangChain, LangGraph, DeepAgents, and LangSmith in particular
* LLM fundamentals: clear, direct prompting with well-structured system messages and context, and tool calling with error handling and retries
* MCP experience: building MCP servers and registering them with an agent as an MCP client
* RAG fundamentals: chunking, embedding, vector stores (Qdrant, pgvector), hybrid queries (BM25), and reciprocal rank fusion, plus the ETL pipelines that feed them
* Agent observability and evals: tracing, single- and multi-turn evals, benchmarking, graders in code, LLM-as-judge, trajectory-based evals, and the data flywheels that connect them (offline and online evals, automations that feed an annotation queue). Nobody has evals fully figured out; we want people who treat them as a first-class problem
* Comfortable with Docker and Docker Compose for local development, and with at least one cloud provider: object storage, managed databases, and deploying a container
* Fluency with AI-assisted development tools (Claude Code, Cursor, Copilot, or similar) as a core part of your workflow, plus good judgment about when to trust them
* A consulting mindset: you listen before you prescribe, you scope realistically, you communicate tradeoffs honestly, and you're comfortable with ambiguity and context-switching across engagements
* Professional-level English (C1 or better), written and spoken. You'll be communicating directly with clients, including non-technical stakeholders, in meetings and in writing
* Working hours that overlap substantially with US Central (Chicago), US Mountain (Denver), US Pacific (San Francisco), or UK (London) business hours
NICE TO HAVE
* Experience with CrewAI, AutoGen, LlamaIndex, or other agent frameworks beyond the LangChain ecosystem
* Production RAG experience at enterprise scale
* Kubernetes (e.g. EKS) and queueing infrastructure (SQS, ElastiCache)
* Experience building agents that support skills, in the coding-agent sense
* Deep eval-driven development experience beyond the basics (custom harnesses, LangSmith evals, EDD as a working style)
* Prior consulting or client-services background
DETAILS
* Rate: $50-100/hr USD depending on experience
* Engagement: contract, project-based, with potential for ongoing work across multiple client engagements
Location: work from anywhere, as long as your working hours substantially overlap one of the time zones above
Possible travel to client sites if needed