LLM & RAG integration
End-to-end RAG pipelines: ingestion, chunking, embeddings, vector databases (LanceDB), reranking, orchestration via LangChain / LangGraph. GPT-4, Gemini or local models (Ollama).
★ Freelance · AI / LLM · Paris
Freelance AI Consultant in Paris
I'm Ali El Mufti, a freelance AI consultant and developer based in Paris. I ship genuinely useful AI features — LLM, RAG, agents, semantic search — into production products, not POCs that never leave the lab. Fullstack (Java, Python, TypeScript): I deliver code that runs, not slides. Remote across Europe or on-site in Paris.
6×
AI search speedup
60s→10s
latency cut
RAG
pipelines in prod
FR/EN/AR
trilingual
What I can own
AI agents & automation
Tool-using agents, function calling, reliable multi-step workflows, guardrails and evaluation. From idea to an agent that actually pulls its weight in production.
Semantic search
Embedding-powered search engines: better relevance, controlled latency, hybrid lexical/vector retrieval. I cut a property search from 60s to 10s at Upfund.
Consulting & shipping
Model selection, cost, security, observability, and clean integration into your Angular/React apps. Fullstack Java Spring Boot or Python FastAPI on the back end.
Concrete results
6×
AI property search — Upfund (Paris)
LLM/RAG-powered search bar on a fintech SaaS platform: response time cut from 60s to 10s, with markedly better relevance.
RAG
RAG pipelines in production
Designed full RAG pipelines (embeddings, LanceDB vector store, reranking) wired into Angular/React frontends, with guardrails and evaluation.
Fullstack
AI wired into a real product
Java Spring Boot / Python FastAPI backends, REST/GraphQL APIs, and AI features shipped into apps that actually run — not abandoned POCs.
Frequently asked questions
What kind of AI projects do you take?
LLM integration, RAG pipelines, AI agents, semantic search, and shipping AI features into existing apps. I focus on real products, not POCs that go nowhere.
Which models and tools do you use?
GPT-4, Gemini, local models via Ollama; LangChain and LangGraph; vector databases (LanceDB); OpenAI and Gemini APIs; prompt engineering and evaluation. On the back end, Java Spring Boot or Python FastAPI.
Can you also build the product around the AI?
Yes — I'm fullstack. I can deliver the AI feature, the backend (Java/Python) and the Angular/React frontend integration, so the whole chain, not just the prompt.
What are your rates and how do I book you?
My day rate starts at €600/day (excl. VAT), adjusted for length, scope and remote/on-site. Book a 30-min call on Collective (app.collective.work/collective/ali-el-mufti) or reach me on Malt — I usually reply within one business day.