Lokesh Gupta

I’m a Data Scientist focused on turning NLP and LLM workflows into production systems that reliably reduce time-to-delivery. I’ve built end-to-end pipelines that combine model optimization with engineering-grade monitoring, logging, and alerting.

At ION Analytics, I designed and deployed an event-driven intelligence generation pipeline on AWS (Lambda, SNS, SQS, Secrets Manager) and Datadog monitoring, automating 300+ intelligence reports and replacing manual analyst workflows—while planning expansion to 10,000+ report generations. I also architected an end-to-end M&A deal automation pipeline using Python, Azure, and PromptFlow, processing 2,000+ deals and reducing turnaround time by ~80%.

I optimize LLM performance with measurable outcomes. I benchmarked and optimized 22 LLM architectures for entity linking across large-scale news datasets, cutting runtime from 55s to 9s (84% reduction) while maintaining production-level accuracy; the solution was adopted across 6 teams.

I enjoy improving systems holistically—accuracy, speed, and operational reliability. By reducing token input by 61% and increasing F1-score by 12%, I improved entity linking performance with production-grade monitoring and alerting, and I bring the same execution mindset to every project I build.