Applied AI Engineer, Hedge Fund

Cedar Peak Partners; Greater London; On-site; GBP 90,000 - 140,000; Full time

Job summary

Cedar Peak Partners in Greater London seeks a hands-on AI/ML engineer to join a global hedge fund client's AI team. You will own end‑to‑end automation, design agents, and deploy in production with audit trails and human-in-the-loop controls.

You will build models for forecasting, classification, ranking, and document intelligence across contracts and meetings, combining entity resolution and knowledge graphs.

Qualifications

• 3+ years shipping production software, ML or AI, with flexibility for builders whose work speaks louder than their years.
• Strong Python and engineering fundamentals, so what you prototype ends up as something the team can build on.
• Real agentic and LLM depth: tool use, orchestration, structured outputs, retrieval, evals, and honest judgment about what an agent can be trusted to do.
• Supervised ML depth, and you know where leakage, overfitting and drift come from and how to catch them.
• The ability to land in an unfamiliar part of the business, get up to speed fast, and be taken seriously by the people who have done the job for years.
• Enough autonomy to embed with a trading team or a finance team and build without waiting for a spec.

Responsibilities

• Agents that run real processes end to end, with the integrations, orchestration, state handling, human checkpoints and audit trail to be trusted in production
• Models that predict and support decisions over years of accumulated firm data: classification, ranking and matching, forecasting, anomaly detection, validated and monitored properly
• Document intelligence across contracts, filings, correspondence and meeting transcripts: extraction, classification, review
• Entity resolution, knowledge graphs and enrichment, with the feedback loops that make a deployed system better over time
• Reusable capability: the second deployment should cost a fraction of the first

Skills

Python

Production software

LLM depth

Supervised ML

Adaptability

Autonomy

Tool use

Orchestration

Job description

Our client is a global hedge fund, and its AI team has a mandate that covers the entire business rather than one desk. Trading, risk, finance, operations, legal, compliance: when a team has a problem worth solving, this is the group that decides the approach and ships it.

The seat

You sit with the team that owns the problem. Most of what you automate has never been written down, so the first job is working out what actually happens, then deciding what an agent should do, what ordinary software should do, and where a person stays in the loop. You own it from that first conversation through architecture, deployment, evaluation and the iteration afterwards, and what works once becomes something the rest of the firm can reuse. No finance background needed. The brief is deliberately broad: production agents against the messy edge cases, not a short list of showcase projects.

What you will build

• Agents that run real processes end to end, with the integrations, orchestration, state handling, human checkpoints and audit trail to be trusted in production
• Models that predict and support decisions over years of accumulated firm data: classification, ranking and matching, forecasting, anomaly detection, validated and monitored properly
• Document intelligence across contracts, filings, correspondence and meeting transcripts: extraction, classification, review
• Entity resolution, knowledge graphs and enrichment, with the feedback loops that make a deployed system better over time
• Reusable capability: the second deployment should cost a fraction of the first

What you bring

• 3+ years shipping production software, ML or AI, with flexibility for builders whose work speaks louder than their years
• Strong Python and engineering fundamentals, so what you prototype ends up as something the team can build on
• Real agentic and LLM depth: tool use, orchestration, structured outputs, retrieval, evals, and honest judgment about what an agent can be trusted to do
• Supervised ML depth, and you know where leakage, overfitting and drift come from and how to catch them
• The ability to land in an unfamiliar part of the business, get up to speed fast, and be taken seriously by the people who have done the job for years
• Enough autonomy to embed with a trading team or a finance team and build without waiting for a spec

Why this seat

• Breadth: a firm-wide mandate, so the problem changes every few months
• Ownership: you are in the room when the problem is described, and accountable for whether the fix gets used