I will fix and optimize rag chatbot retrieval and hallucinations

M
mohammadaminb
M
mohammadaminb
Amin bm
Sommige informatie wordt in het Engels weergegeven.

Over deze dienst

Does your RAG chatbot retrieve weak evidence, cite the wrong source or underperform as your data grows?


I diagnose, fix and optimize existing RAG chatbots that miss relevant passages, rank evidence poorly or answer beyond their sources.


I inspect ingestion, chunking, metadata, lexical and vector search, hybrid fusion, ranking, context construction, citation mapping and insufficient evidence behavior. You receive reproducible test cases, prioritized findings and clear before and after results. Standard and Premium include agreed code changes.


My production retrieval backend combines full text and semantic search with weighted rank fusion, source deduplication, evidence reading, caching and regression checks. When appropriate, I can optimize basic vector retrieval into a better tested hybrid pipeline. I focus on measurable retrieval accuracy, not prompt changes alone.


I work with Python or JavaScript, REST APIs, FastAPI, Cloudflare Workers, PostgreSQL pgvector, Qdrant, SQLite FTS5 and comparable stacks.


Message me before ordering with your stack, failure examples, access method and 2 to 5 real questions.

Maak kennis met Amin bm

Amin bm

Programming and Tech

  • Afkomstig uitVerenigd Koninkrijk
  • Lid sindsaug 2026
  • Talen

    Engels
Python developer with a Master's in Economics who builds complete systems, not just scripts: AI-powered search and RAG backends, desktop applications, document-processing pipelines, research tools and API integrations. Two of my products are public: a 37,000-line production retrieval backend with a live demo, and an open-source desktop app that turns whole books into AI-ready structured text. Strong mathematics core - ranking algorithms, deduplication, capacity planning. Documented, tested code, honest scoping, available after delivery.