A collection of projects spanning AI engineering, data science, and full-stack development
AI Inbound Calling Agent
Architected and led a four-person team delivering a voice-driven inbound calling agent. Processes calls through Google STT → FastAPI → RAG pipeline (OpenAI embeddings + pgvector) → GPT-4o Mini → Google TTS, with natural Urdu-language intent classification at production scale.
Real-Time Chat Application
Deployed a production real-time messaging app on a MERN stack with Firebase Authentication for JWT-based identity and Socket.io for sub-second message delivery, achieving zero-polling live presence tracking across concurrent users.
Computer Vision · 95%+ Accuracy
Achieved 95%+ face classification accuracy on a binary real/fake dataset by fine-tuning XceptionNet (CNN) on augmented training data. Transfer learning reduced training time by 60% versus training from scratch.
End-to-End Data Engineering on GCP
Processed and modelled an Uber-scale trip dataset end-to-end — raw ingestion → dimensional schema design → orchestrated ETL via Mage → analytical dashboard deployed on GCP — reducing query time on aggregated data by 40% over flat-file analysis.