About

Senior AI Engineer (RAG + GPT) to Build Evidence-Based Insight Engine for Health App Overview We are building an evidence-based AI insight engine for Elli Cares, a senior health & wellbeing platform used by older adults and their families. The goal is to generate safe, non-diagnostic, evidence-backed wellbeing insights based on structured inputs such as: - symptoms - medication adherence - hydration - mood - sleep & mobility signals - behavioural changes This is not a chatbot project. It is a production-grade RAG system with strong safety constraints and hallucination prevention. What You’ll Build You will design and implement the Elli Evidence Engine (EEE) — a backend AI service that: Uses Retrieval-Augmented Generation (RAG) with a curated medical evidence base Generates non-diagnostic, senior-friendly insight summaries Enforces strict safety rules (no diagnosis, no treatment advice, no dosages) Classifies insights by severity (info / monitor / see GP / urgent) Returns structured JSON outputs for use in a Flutter app and webview integrations Our existing developers will handle: Flutter app integration Webview / partner integrations (e.g. patient portals) PDF generation for GP reports You will focus purely on the AI + RAG backend. Core Responsibilities Design and implement a RAG pipeline (retrieval → reasoning → summarisation) Build and manage a vector database (e.g. Pinecone, Weaviate, Chroma, Milvus) Ingest and chunk curated evidence sources (medication safety, symptom guidance, dementia resources, red-flag guidance) Implement hallucination prevention and safety guardrails Produce structured outputs (short summary, full summary, severity, drivers) Expose clean backend APIs for insight generation and retrieval Add logging/traceability for auditability and tuning Required Skills (Non-Negotiable) Proven experience building RAG systems in production Strong experience with LLMs (OpenAI GPT-4/5 or equivalent) Hands-on experience with vector databases Strong Python backend skills (FastAPI preferred) Experience implementing LLM safety constraints and guardrails Ability to build evidence-constrained generation (not free-form chat) Experience designing structured JSON outputs from LLMs Strongly Preferred Experience in healthcare, legal, fintech, or other regulated domains Experience building zero-hallucination or citation-based RAG systems Familiarity with medical or scientific content summarisation Experience with LangChain, LangGraph, LlamaIndex, or similar orchestration frameworks What This Is NOT ❌ A chatbot ❌ Prompt-only work ❌ No-code / low-code AI ❌ Frontend or Flutter work ❌ Medical diagnosis system Engagement Details Duration: expected 6 - 8 weeks Estimated hours: expected 130hrs - 190hrs total Rate: Open (we value expertise over lowest cost) Timezone: Flexible Communication: Slack + weekly check-ins How We’ll Evaluate You Please include in your proposal: A brief description of a RAG system you’ve built (especially where hallucination prevention mattered) Which vector DBs you’ve worked with How you typically constrain LLMs to evidence Any experience in healthcare, legal, or safety-critical AI systems Why This Project Is Interesting You’ll be building the core intelligence layer of a real, growing product The system will be used by older adults and families — real-world impact Strong emphasis on AI safety and quality, not hype Opportunity for ongoing work as the platform scales