Junior AI Engineer

We are building a premium AI consultancy targeting an underserved whitespace: transforming small-to-mid-cap portfolio companies and finance firms that cannot afford to get AI wrong. We hire only top-tier talent. Our 15-person engineering team ships agentic systems, retrieval-augmented generation (RAG) pipelines, custom copilots, and document automation into clients including Nviya-Prime, Maoki, Europol, NYPL, J.P. Morgan, and Intel x Activeloop.

This role is for an early-career engineer who wants to grow inside a serious AI engineering team. You will work alongside our senior engineers on real production LLM systems and learn the full engineering loop. The role is fully internal by design: no client calls, no executive conversations, no on-site work. Senior engineers and the deployment strategist own all of that. Your job is to build, evaluate, and ship. Over time, as you grow into a senior engineer, there is a clear path into more client-facing or forward-deployed work.

• -Build and maintain components inside production LLM applications: retrieval pipelines, agent workflows, eval harnesses, integrations, internal tools.
• -Own well-scoped tickets end-to-end: design, implementation, tests, evals, code review, deployment.
• -Implement and improve RAG pipelines: parsing, chunking, indexing, retrieval, reranking, grounding, source attribution.
• -Build and maintain agent components: tool calling, structured outputs, orchestration steps, retries, human review hooks.
• -Build evaluation suites: golden datasets, regression tests, retrieval and generation metrics, observability hooks.
• -Integrate AI systems with client data sources, document repositories, customer relationship management (CRM) systems, knowledge bases, internal application programming interfaces (APIs).
• -Write clean, tested, documented Python; contribute to code review and engineering discussion.
• -Contribute to reusable firm assets: prompt libraries, agent skills, eval harnesses, reference patterns, internal documentation.

You should already be a daily, native user of agentic coding tools: Claude Code, Codex, Cursor, or similar. Not occasional experiments. We expect them in your default workflow for planning, codebase exploration, implementation, refactoring, testing, debugging, and documentation. We hire on this.

The standard is intelligent supervision, not blind delegation. You give the agent the right context, constrain the task, inspect what it produces, run the tests, and catch weak assumptions before committing. You never ship code you have not understood.

Beyond coding tools, you will help build reusable AI infrastructure the team uses across engagements: Claude Skills, Model Context Protocol (MCP) servers, sub-agents, eval harnesses, prompt patterns, workflow templates.

• -1-3 years professional software engineering experience, or equivalent strong evidence (open-source work, published projects, hackathons, technical writing).
• -At least one shipped LLM project beyond simple chat: a working RAG pipeline, an agent with real tool use, an internal copilot, a document analysis system, or comparable.
• -Strong Python. Comfort with APIs, JSON, async, packaging, testing.
• -Working knowledge of OpenAI, Anthropic, Gemini, or comparable LLM APIs.
• -Practical understanding of RAG: embeddings, chunking, vector search, reranking, basic retrieval evaluation.
• -Comfort with Git, Docker, basic cloud deployment, Linux command line.
• -Habit of evaluating LLM outputs rather than trusting them: written test cases, golden sets, regression checks.
• -Daily use of agentic coding tools (Claude Code, Codex, Cursor) for real work, with careful inspection of what they produce.
• -Strong written communication. Clear pull request (PR) descriptions, clear questions, clear documentation.

• -Familiarity with LangChain, LangGraph, LlamaIndex, LangSmith, Langfuse, Braintrust, MCP servers, or OpenAI Agents SDK.
• -Vector databases such as Pinecone, Weaviate, Qdrant, Chroma, or pgvector.
• -TypeScript, React, Next.js, Postgres, or Supabase.
• -Domain interest in finance, investment, operations, sales, or customer support.
• -Experience with fine-tuning, distillation, or open-weight models.

• -How to design, build, evaluate, and ship production LLM systems, not demos.
• -How to debug agent failures, fix retrieval quality issues, and harden systems for real use.
• -How to evaluate models, prompts, and pipelines with discipline rather than vibes.
• -How to use agentic coding tools at a senior level: planning, supervision, intelligent rejection.
• -How AI engineering translates into real client value, observed through the senior engineers and strategists you support.

An early-career engineer with strong technical fundamentals, real curiosity about AI engineering as a craft, and the temperament for a high-bar environment. You care about correctness, evaluation, and useful outputs more than impressive vocabulary. You read other people's code carefully, ask sharp questions, ship fast under intelligent supervision of agentic tools, and grow fast.

Towards AI is the AI deployment and education firm, founded in 2019: 500,000+ AI practitioners taught, 200,000+ newsletter subscribers, 100,000+ community members, and 10,000+ copies sold of our O'Reilly book Building LLMs for Production. Our Towards AI Deployment practice runs a specialist team of 15+ AI engineers focused on investment firms, PE, portfolio companies, and finance firms in regulated industries.

Co-founded by Louie Peters (ex-J.P. Morgan VP, credit research) and Louis-Francois Bouchard (ex-Mila, Polytechnique Montreal). Our engineers grow up through our own community and Learning programmes into Deployment pods, so you would work with people who stay at the forefront of the field. In a pod you own a real system end to end, in production, from day one.