Senior AI (LLM) Engineer

Impressit is looking for a Middle/Senior AI (LLM) Engineer to join a UK-based product team building an AI-powered SaaS platform for property and facilities management. The product uses conversational AI and automation to streamline customer communication and operational workflows.
This is a great opportunity to work on production AI/LLM and agentic functionality as an embedded member of an established engineering team. You will continue and evolve an existing AI stack, working with production code and integrations rather than building a solution from scratch.

— Continue the existing AI/LLM and agentic development work previously owned by the core engineer
— Develop, maintain, and optimize production AI workflows and agent architectures using Python (FastAPI, async/concurrency)
— Work with LangChain / LangSmith for AI workflows, tracing, observability, and systematic evaluation
— Design and enforce secure data access patterns for LLM tools (least-privilege access, role-based permission checks outside model prompts)
— Integrate AI services with the core platform, databases, microservices, and messaging channels (Microsoft Bot Framework, WhatsApp, potential voice integrations)
— Work alongside TypeScript microservices, reading and understanding existing code, APIs, and service-to-service integrations
— Take ownership of an existing production codebase, debug complex concurrency and model integration issues, and ensure system reliability
— Design, implement, and maintain rigorous LLM evaluation pipelines to measure model behavior, prevent regressions, and catch production failure modes
— Participate fully in the client’s engineering practices (standups, code reviews, architectural discussions, retrospectives) and reason through engineering trade-offs with the team

— 3+ years of commercial software development experience with strong commercial mastery of Python
— Deep, practical understanding of asynchronous Python (asyncio, event loop, handling concurrency in frameworks like FastAPI) in concrete production scenarios
— Hands-on production experience building and scaling LLM-powered applications, conversational AI, or AI agents
— Proven track record in LLM Security: practical experience controlling data access, implementing authorization checks at the system level (beyond relying solely on prompt instructions or pattern filtering), and mitigating prompt injection risks
— Practical experience with LLM evaluation & testing: ability to build, measure, and leverage evals to catch real-world edge cases and regressions
— Experience working with LangChain, LangSmith, or alternative orchestration and observability toolkits
— Ability to read and navigate TypeScript microservices and integrate with internal APIs
— Experience stepping into an existing codebase, taking ownership, and evolving software without unnecessary rewrites
— Strong problem-solving skills with the ability to discuss complex engineering scenarios, architecture decisions, and real-world trade-offs live
— Advanced level of English (both written and spoken) for daily technical discussions with UK-based engineering leadership

— Hands-on experience with messaging channels: Microsoft Bot Framework, WhatsApp, or voice integrations
— Experience working with distributed TypeScript / Python microservice ecosystems
— Deep knowledge of evaluation methodologies (LLM-as-a-judge, regression testing, dataset creation for domain-specific tasks)
— Practical exposure to Claude Code, Cursor, or similar AI-assisted development tooling
— Domain experience in PropTech, property/facilities management, or enterprise SaaS
— Experience designing fallback mechanisms and deterministic guardrails for probabilistic agent workflows