Artificial Intelligence Engineer

Role & responsibilities :

Design, build, and ship agentic workflows for Weave: orchestration, tool selection, retries, failure handling, and the handoffs between LLM reasoning and deterministic code.

Build and extend the tool layer, including our MCP server, designing tool interfaces and schemas that models use correctly under real conditions.

Engineer against hallucination: structure pipelines so the agent retrieves and reasons but never invents a number, with verification and traceability built in.

Build evaluation harnesses and golden datasets, and hold new work to them. Regression testing is a first-class part of the job here, not an afterthought.

Work with LLMs across the stack, including prompt and context engineering, retrieval, structured outputs, and fine-tuning where it is genuinely the right tool.

Build and maintain the data pipelines, databases, and APIs that Weave and our other platforms depend on.

Optimise for the constraints that actually bite: latency, token cost, and reliability across long multi-step runs.

Support deployment, testing, monitoring, and observability of AI systems in production, and diagnose failures when they happen.

Work directly with our analysts and product team to understand what a correct model or answer looks like, and encode that judgement into the system.

Prototype and evaluate new AI tools, frameworks, and techniques, and document what you find clearly enough that the team can act on it.

Preferred candidate profile

[2+ years] of professional experience in software engineering, AI/ML, or data engineering, with strong Python and backend engineering fundamentals, including asynchronous processing and streaming workflows. You write production code, not only notebooks.

Demonstrated hands-on experience delivering real Agentic AI projects, meaning agents that plan, make decisions, invoke tools, manage context, and complete multi-step workflows end to end. You have seen these systems fail in production and understand why.

Practical fluency in prompt and context engineering: structuring context, decomposing tasks, and knowing when a prompt is the wrong solution.

Practical experience with agent orchestration patterns such as routing, delegation between agents, memory, structured outputs, and human-in-the-loop workflows.

Solid grounding in machine learning and AI fundamentals, and comfort with libraries such as NumPy, Pandas, and PyTorch or TensorFlow.

Experience integrating agents with production APIs, databases, enterprise applications, and external services.

An empirical instinct: you reach for an evaluation before an opinion, and you are willing to say a system is not ready.

Experience implementing guardrails, permissions, and secure tool execution, plus failure handling, observability, and agent evaluations.

Clear written and verbal communication, including with people who are not engineers.

Familiarity with Agentic AI technologies such as MCP, Gemini, the OpenAI Agents SDK, LangGraph, Anthropic's tooling, or comparable frameworks.

Ability to walk us through your personal contributions to at least one substantial or production Agentic AI implementation, covering its architecture, tools, safeguards, challenges, evaluation approach, and measurable results.

Note on scope: basic chatbots, prompt engineering alone, simple RAG pipelines, tutorials, certifications, and thin wrapper applications are not considered sufficient Agentic AI experience.

GOOD TO HAVE

Contributions to open-source agent tooling, MCP servers, or model-facing tool interfaces used by others.

Experience with financial data, spreadsheets, Excel automation (openpyxl, Office.js), or document-analysis workflows.

Background in finance, fintech, or financial data, so you know what consensus, guidance, and a segment KPI are, and why close enough is worthless in this domain.

Depth in LLM tracing and evaluation infrastructure, building the harnesses and golden datasets rather than just reading the dashboards.

Applied ML or research background: forecasting, time-series modelling, or information extraction from filings and unstructured documents.

Cloud deployment and CI/CD experience.