Muhammad Ahmad Mujtaba

I've built end-to-end AI systems for international clients, from raw data ingestion through deployed services. My work spans LLM fine-tuning, RAG pipelines, forecasting, computer vision, and full-stack ML backends.

At HKS TechLabs, I'm the sole ML engineer building an AI pricing recommender from Supabase ingestion through SKU classification and STL decomposition. I developed LightGBM demand forecasting with Tweedie and quantile regression, elasticity simulations, and a two-phase LP and SLSQP price optimizer with per-SKU explainability.

For a SaaS startup, I built a taxonomy-driven RAG recommendation system with Qdrant that reduced short-query times from 15–20 seconds to 5–7 seconds and larger-query times from 50–60 seconds to 10–13 seconds. I also migrated it from a monolith supporting 4–6 concurrent users to async microservices supporting 40–50 users.

At EnDevSols, I trained and deployed computer vision, NLP, and generative AI models, maintained automated training pipelines, and built APIs with FastAPI. I hold a B.Sc. in Artificial Intelligence from COMSATS University Islamabad.