At TekkDev, I architected and deployed the company’s production LLM platform solo on an RTX 6000, then became ML/AI Team Lead. I serve Mistral 7B AWQ through vLLM, built a bilingual real-time voice pipeline, and added semantic caching, prompt-injection defense, and rate limiting.
Previously at Xorbix Technologies, I owned end-to-end ML pipelines for manufacturing and fintech clients on Databricks. I built an LLM evaluation framework that improved RAG response quality by 24% and optimized FAISS vector search to increase query speeds by 15%.
I also fine-tuned text-to-SQL models, shipped forecasting and anomaly-detection systems, and build practical RAG applications such as a Pakistani legal chatbot.
