Shabbir B.

I build AI agents, RAG chatbots and LLM features for products that real users depend on. Chatbots that answer from your own documents and cite the source, agents that call tools without looping, and AI workflows that still work a month after launch. A lot of my work starts with an AI app that looked fine in testing and then broke on real data: retrieval pulling the wrong chunks, the model answering outside the knowledge base, an n8n flow failing silently between steps. I track down why and fix it. When the project is new, I build it from scratch, anywhere from a quick proof of concept to a production SaaS feature. ✦ What I build: » AI agents and agentic workflows: tool calling, memory, routing between agents, and human approval steps where a wrong action would be costly. » RAG systems and knowledge base chatbots: document ingestion, chunking, embeddings, vector search, source citations, and a clear "I don't know" when the answer isn't in your content. » LLM integration into existing apps and SaaS products: OpenAI, Claude and Gemini APIs, structured outputs, fallbacks and retries. » AI workflow automation with n8n, Make and Zapier, connected to your CRM, inbox, Slack or database. » Full stack AI apps: backend APIs and React or Next.js frontends, from MVP to multi-tenant SaaS. ✦ Work I've shipped: » A multi agent AI platform where agents are assigned by role and hand work to each other to run real business workflows. » AI features for a large US ecommerce platform: fraud detection, product description generation, ad performance analysis and image generation pipelines. » An AI ops assistant for a 30 year old insurance consultancy. Staff get answers grounded only in approved SOPs, and compliance guidelines are generated per agency. Audit handling time dropped 45%. » An n8n email workflow that sorts incoming mail by type, drafts AI replies in the right tone, logs everything and alerts the team on Slack. Nobody triages the inbox by hand anymore. » Patient intake and therapy adherence systems for a NASDAQ listed healthcare company, used by millions of patients across 40+ US states, plus a hybrid care outpatient platform with an AI chatbot and symptom tracking. ✦ Stack: » AI: OpenAI, Claude, Gemini, LangChain, LlamaIndex, RAG, embeddings, prompt engineering, fine tuning » Vector databases: Pinecone, Weaviate, FAISS, Supabase » Automation: n8n, Make, Zapier, Pipedream » Backend: Python, Node.js, Laravel, REST and GraphQL APIs » Frontend: React, Next.js If you have an AI feature that works in a demo but not on your real data, or you're starting one and want it built for real usage, send me the details. I'll reply with how I'd approach it and what I'd check first.