Nagraj Deshmukh

• Led NLP-driven NPS and call analytics pipelines — insights drove 12% YoY reduction in member care calls
• Proactively led GPT-3.5 → GPT-4o migration across teams, preventing downstream production failures
• Designed LLM evaluation framework using DeepEval to monitor performance and regressions across enterprise AI
• Winner, Alliant Innovation Challenge for Underwriting AI Assistant concept

• Built ETL copilot using transformer models to extract unstructured data from financial reports — improved data accessibility by 70%

• Engineered an autonomous LLM marketing agent that boosted post frequency 40% and engagement 25%
• Built prompt-chaining and RAG pipeline using open-source LLMs, achieving semantic alignment score of 0.8
• Presented at INFORMS Analytics Conference 2024; co-authored paper at International Conference on Data Science

• Optimized pose detection models — 200% speed improvement, 65% accuracy gain
• Led exercise categorization project that scaled the product from 5 to 400+ exercises
• Directed 4-person DS team on RTM tool that secured $7M in funding

• Delivered Credit Risk Monitoring application used by 1,100+ clients
• Resolved 6 categories of OWASP security vulnerabilities
• Led C++ → Java module migration in 2 months; built documentation pipeline cutting team response time 30%

• Developed metaheuristic optimization algorithms for high-dimensional problems
• Published 2 papers (20 citations) in Evolutionary Intelligence (Springer) and Knowledge Based Systems (Elsevier)

NLP-driven pipeline analyzing member feedback and call transcripts at Alliant Credit Union. Drove 12% YoY reduction in member care call volume.

Enterprise AI monitoring system using DeepEval to track performance regressions across GPT-4o deployments. Enables automated quality gates for AI outputs.

Transformer-based system to extract structured data from unstructured financial reports. Replaced manual extraction workflows at Nedl Labs.

Optimized MediaPipe + LightGBM pipeline for real-time exercise categorization. Helped scale Spry Therapeutics from 5 to 400+ exercises, supporting a $7M funding round.