AI Engineer

What matters most

• Build enterprise AI workflows

  The role builds AI-powered products and internal tools using LLMs, retrieval systems, agents, and structured outputs across client and internal projects.

• Move prototypes into dependable delivery

  Makai Labs calls for rapid prototyping and iteration, followed by reliable, client-ready systems that meet client needs.

• Evaluate approaches, not just implement them

  Research, benchmarking, testing, and evaluation of new approaches are explicit responsibilities; experience running AI experiments is part of the ideal background.

• Own scoped work and make it reusable

  The engineer owns smaller projects or sub-projects with team guidance, contributes to shared platform tooling, and maintains clear internal documentation while working with engineering and design.

Likely challenges to think through

• Speed versus reliability

  Preparation hypothesis: fast experiments may need substantial changes before they are client-ready. Prepare an example showing how you selected a prototype, tested its behavior, and addressed failures before delivery.

• Choosing among AI approaches

  Preparation hypothesis: LLM, retrieval, and agent workflows can differ in complexity and predictability across domains. Prepare to explain how you would benchmark alternatives against a defined task rather than adopt a newer approach by default.

• Balancing client-specific work with reuse

  Preparation hypothesis: varied client processes may make shared platform work difficult to generalize. Prepare an example of deciding what to keep specific to a deliverable, what to turn into reusable tooling, and what to document for teammates.

Topics to prepare

• LLM workflow design

  Review when to use retrieval, an agent, or a structured output; sketch a workflow for one enterprise automation task and explain each choice. The posting explicitly calls for workflows using all four.

• Retrieval and search concepts

  Compare how document selection, embeddings, and search affect the information supplied to an LLM; walk through two cases where retrieval might return the wrong material. Retrieval is a responsibility, while embeddings and search are nice-to-haves.

• AI evaluation and iteration

  Define success criteria for a small AI feature, then outline a benchmark and how you would compare two iterations. The role explicitly includes experiments, benchmarking, testing, and evaluation.

• Prototype-to-client-ready tradeoffs

  Take a rough AI prototype and identify its failure cases, testing needs, and reusable components. This prepares for the stated work of improving prototypes into reliable systems and contributing shared tooling.

Skills to revise

• Build a Python AI workflow

  Implement a small Python workflow that accepts an input, calls an LLM, validates a structured result, and handles an invalid result. Strong Python skills and LLM workflows with structured outputs are explicit in the posting.

• Test and benchmark an iteration

  Create a small set of representative inputs, run two versions of a workflow, record outcomes, and explain the change you would keep. This practices the advertised experimentation, testing, benchmarking, and rapid iteration.

• Prototype a retrieval-backed feature

  Build a minimal document lookup step and inspect what it returns before using the result in an LLM workflow. The role explicitly involves retrieval systems; document processing and PDF workflows are only nice-to-haves.

• Document a handoff

  Write a brief note for a small AI component covering its purpose, how to run it, known limitations, and next checks. Clear internal documentation and ownership of smaller deliverables are stated responsibilities.

• Expose a workflow through an API

  As optional preparation, wrap a small Python workflow in an API endpoint and test a successful request and an error case. APIs and backend frameworks such as FastAPI or Flask are listed as nice-to-haves, not requirements.

• Select an AI application to walk through

  The role builds AI-powered products and internal tools. Prepare a concise walkthrough of a project from coursework, research, an internship, or personal work: the problem, your contribution, implementation, results, and what you would change.

• Show evidence of Python implementation

  Strong Python skills are listed as ideal experience. Identify code you can explain, including how you structured it, tested it, and handled a failure or edge case.

• Practice an enterprise workflow design

  As a preparation exercise, sketch how you would turn an operational task into an AI workflow. Explain where an LLM, retrieval, an agent, or structured output would help—and where a simpler approach would suffice.

• Prepare an experiment and evaluation story

  The posting emphasizes rapid iteration, benchmarking, testing, and evaluation. Bring one example with a baseline, what you changed, how you measured the result, and the decision the evidence supported.

• Explain the path from prototype to delivery

  Prepare a short plan for making a promising prototype reliable and client-ready, covering tests, failure handling, clear internal documentation, and any reusable tooling you would extract. These are preparation points, not a stated company process.

• Choose relevant optional technical examples

  If you have them, select evidence involving APIs or backend services, retrieval and search, document or PDF processing, or local development. The posting lists these as nice-to-haves; focus on work you can explain rather than trying to cover every tool.

• Write questions about ownership and collaboration

  Bring questions about how smaller projects receive team guidance, how engineering and design coordinate on client needs, how approaches are evaluated, and what Eastern-time collaboration looks like for this remote contract role.