SADIQ SHEHU MUSA

Architected and deployed a full production AI system, currently in active use, to autonomously qualify leads and drive sales conversations via WhatsApp for a real estate client.

Developed an intent detection engine, multi-phase conversation state machine, and objection-handling logic, dynamically adapting bot behavior by tracking `pressure_level` and `objection_type` in Supabase.

Designed and shipped an end-to-end ML pipeline, from webhook event data ingestion to state classification, LLM-powered response generation, and structured Supabase logging.

Resolved critical production issues, including Pydantic schema mismatches, webhook 401 auth failures, and bot-blocking `ai_paused` status logic, ensuring system stability and reliability.

Managed infrastructure spanning Railway (Evolution API), Render (FastAPI), Supabase (Postgres + state store), and Vercel (frontend dashboard) for robust deployment.

Built a full-stack adaptive study application for university students, detecting subject categories from uploaded material and applying evidence-based teaching strategies per NUC faculty guidelines.

Implemented a robust RAG pipeline, including PDF ingestion (PyMuPDF), chunk embedding (sentence-transformers), vector storage (Qdrant), and contextual response generation (Groq LLAMA 3).

Deployed FastAPI backend on cloud with Supabase auth integration, optimizing for resource-constrained environments by pinning NumPy and installing CPU-only PyTorch.

Designed for 8 NUC-aligned faculty categories, establishing it as the first AI study tool specifically tailored to the Nigerian university curriculum structure.

Developed a retrieval-augmented generation (RAG) system for document summarization and contextual Q&A over user-uploaded files, enhancing information access.

Implemented a comprehensive pipeline: document ingestion, text chunking, embedding generation, vector retrieval, and response synthesis for efficient data processing.

Designed a scalable architecture capable of supporting multiple document formats and concurrent user sessions, ensuring high availability and performance.