AI Engineer
AI Engineer | Python | LLM | Agentic Workflows | LangGraph | EducationTech | Remote, London | £65k Looking to help shape AI Strategy for a scaling Tech-for-Good business I'm partnered with a AI-first data and analytics platform built for schools, multi-academy trusts and education groups. Their technology connects fragmented school data - attendance, behaviour, wellbeing, assessment, SEND and unstructured documents - and uses a multi-agent AI system to surface actionable insights for school leaders in seconds.They are in a high-growth phase, backed by Innovate UK funding, with a roadmap spanning predictive analytics, supplier intelligence, and international expansion. The team is mission-driven, remote-first, and growing quickly - every hire has very meaningful impact.They are looking for an AI Developer to extend, improve and scale their 22-agent agentic AI platform. The system is built on LangGraph with a supervisor-of-supervisors architecture, integrates multiple LLMs (including Google Gemini), and is monitored via LangSmith.You’ll build and refine agent sub-graphs, improve prompt engineering, contribute to LLM benchmarking, and help evolve the platform as new capabilities are added — including semantic query generation and predictive analytics integration.This is a production engineering role: you’ll write robust Python code, ship features, and work in a cloud-native environment.Core Responsibilities:
- Agentic Workflow Development: Build and refine LangGraph/LangChain agent sub-graphs and multi-step workflows.
- LLM Integration: Improve prompt engineering, orchestrate multiple models, and implement tool-calling logic.
- System Observability: Use LangSmith or equivalent tracing/logging to monitor agent behaviour and performance.
- Benchmarking & Evaluation: Contribute to LLM benchmarking across tasks and model families.
- API Engineering: Build and consume REST APIs; integrate agentic workflows with backend services.
- Architecture Evolution: Help evolve the platform as new capabilities (semantic query generation, predictive analytics) are added.