Varun Totakura

About Varun Totakura

Passionate AI/ML Engineer with 5 years of professional experience. I thrive on innovation…

Experience

1. Ai Engineer

   EXL

   Oct 2024 — Present · NY, US

   Researched and developed a machine learning model using Group Classification for Deep Imbalance Regression to address skewed data, boosting performance from 20% to 70%. • Built End-to-End ML models, ensuring model stability and fairness by implementing bias checks & mitigations, and deployed data drift detection mechanisms to prevent future drifts. • Worked with diverse healthcare data sources, including claims, SDOH, medical notes, membership records, and chronic disease datasets having 100+ million records. • Utilized Generative AI models such as Claude to extract features from medical notes, integrated them with tabular data, and achieved a 10% performance improvement. • Led a team of 4 by planning tasks, delegating responsibilities, providing technical support, and ensuring on-time delivery of client projects. • Developed a Generative AI solution for fax data extraction using OCR, integrating a RAG system with OpenAI GPT-4o to deliver context-driven answers with 20% improved response relevance. • Engineered 15 CI/CD ML pipelines deployment using Jenkins and Apache Airflow DAGs on GCP, launching DataProc clusters to process data and generate predictions, with results stored in BigQuery tables for downstream process. • Optimized text processing code by parallelizing tasks with Dask and Multi-threading, reducing execution time from 4 hours to under 2 hours while maintaining accuracy. • Productionized two Kubernetes-based applications, optimizing container orchestration for scalability and achieving good uptime across distributed healthcare systems. • Automated deployment of local Python scripts to GCP-compatible Airflow workflows, slashing deployment time from over 24 hours to under 30 seconds, boosting team productivity by 50%. • Refactored and optimized a data pipeline processing 1M+ records daily, enhancing code readability and accelerating execution from 9 hours to under 5 minutes through efficient resource utilization.

Education

• Florida State University

  Master of Science - MS, Computer Science

• Jawaharlal Nehru Technological University

  Bachelor of Technology, Computer Science

  2016 — 2020