Likhith Vennam

Data Scientist responsible for developing and fine-tuning large language models to power a spend-analysis chatbot, cutting search time by 70%. Built ML models to predict optimal buy cost and supplier for silicon procurement. Architected and deployed end-to-end ETL pipelines processing 500+ jobs daily from multiple sources into Snowflake and MySQL, reducing data latency by 50%. Led classification model development for BOM parts and tiers; implemented CRF, BERT, and REGEX-based pipelines to detect and redact PII and PHI across four data sources with 90% precision. Designed prediction models for In-Store Repair and Mail-In Turnaround Time, and deployed APIs on Apple's Shuri platform (UAT and Production). Maintained IRIS, ISR, and Mail-In APIs; developed a model to forecast Repair Templates; led non-inclusive language replacement initiative within the developer ecosystem.

Led an ML solution for request tracking to automate routing and responses using topic modeling and auto text generation, reducing response time from 48 hours to 4 hours.

Analyzed classification and clustering techniques on wireless surveillance sensor motes to determine intrusion types (Visual, Magnetic, Acoustic, Thermal, Vibration); performed rigorous data analysis to define intrusion characteristics.