Job Specifications
Data Scientist / AI Engineer
Real-time Payment Context (Fintech) | London | Contract | Outside IR35
In short:
A high-growth fintech is looking for a contract Data Scientist / AI Engineer to help build and ship production-grade scam intelligence layer. You’ll work across data, modelling, and deployment - turning multi-source signals into reliable, explainable risk decisions under real-world constraints like latency, uptime, and auditability.
About the company:
The company is building a payment intelligence layer for banks - running real-time “investigations” on payments to provide rich context on the counterparty and situation. The goal: intercept scams while ensuring genuine payments flow smoothly. They’re early-stage, moving fast, and operating in a domain where correctness, security and reliability are non-negotiable.
Who we’re looking for:
You’re a hands-on contractor who can get productive quickly, operate with minimal oversight, and deliver in production. You’re comfortable owning the full loop: data → modelling → deployment → monitoring → iteration, and you care about building systems that are practical, explainable, and bank-grade.
What you’ll do:
Build and ship scam risk models and signals (typology classification, risk scoring, decision logic)
Engineer features across heterogeneous data: transaction context, behavioural sequences, counterparty signals, network/graph patterns, and unstructured evidence
Design calibrated outputs (scores + reason codes) that are actionable and explainable for banking workflows
Own evaluation end-to-end: leakage avoidance, cost-sensitive metrics, thresholding, phased rollouts, and post-incident learning
Productionise ML: packaging, deployment, monitoring, drift detection, and retraining strategies
Partner with backend/product teams to integrate intelligence into real-time payment flows
(Where useful) support agent/LLM workflows for evidence gathering and synthesis — while keeping the decision core predictable and auditable
Must-haves:
Strong experience shipping applied ML into production (not just experimentation)
Strong Python + ability to write maintainable, tested code
Strong SQL + comfort working directly with messy, high-volume data
Solid modelling judgement: calibration, leakage, bias, thresholding, cost trade-offs, monitoring/drift
Experience operating in environments where reliability, latency, and explainability matter
Able to work autonomously and communicate progress clearly in a fast-moving team
Nice-to-haves:
Experience in fraud/scams, payments, risk, trust & safety, AML, or adjacent domains
Familiarity with graph/network features and entity resolution style problems
MLOps tooling exposure (model registry/MLflow, feature stores, orchestration)
Cloud-native/event-driven system familiarity and comfort collaborating with platform/backend engineers
Experience integrating unstructured signals (text/embeddings/RAG-style pipelines) into decision systems
Why this contract:
High-impact work: stopping scams before money leaves
Real-time, bank-grade ML problems
Autonomy + speed - you’ll ship meaningful changes quickly