Kumail R.

I build reliable data and analytics foundations and work on efficient LLM deployment. My current quantization project explores post training quantization and quantization aware training, including weight and activation formats such as FP8, FP4, MXFP4 and NVFP4, plus KV cache quantization for long context inference. I evaluate model quality, memory use, latency, throughput and target hardware compatibility. My data and analytics work spans Python, SQL, Azure, Microsoft Fabric, Databricks, Snowflake, dbt and Power BI. I help teams improve data reliability, analytics and AI readiness, connecting technical choices to measurable business outcomes. If you’re building efficient AI systems or trusted analytics, connect with me and tell me what you’re working on. Recent work includes a DistilBERT NER project: