Hello, I'm vaibhav
I’m a Senior Applied Scientist with 14 years of expertise in AI/ML, specializing in Computer Vision, NLP, and Generative AI. My current passion lies in Agentic AI, where I’m building innovative solutions to explore its transformative potential.
experience
PG&E (Pacific Gas and Electric) | Computer Vision Scientist | Remote - Seattle Nov 2021 -Present
• (CV) To mitigate wildfire risks, I developed and deployed Computer Vision models (ResNet, Faster RCNN/mmdet) to improve issue identification accuracy, achieving a 9% improvement in wildfire management.
• (CV) Scaling issue identification into identifying different insulator types, materials and color, reducing ground inspector’s time by 90% in completing the transmission line’s inspection work.
• (CV Research) Revamping old models to newer architecture designs via network layer modifications, strategic weight freezes, batch normalization, and optimized parameters, improving precision by 27% on average across all the models.
• (MLOps) To process large volumes of images in real-time, I built scalable CV pipelines using PyTorch, ONNX, Seldon Core, Docker, Kubernetes, and Jenkins, enabling real-time processing of 500K images/hour and streamlining operations.
• (MLOps) To address high training costs, I designed a Sagemaker training framework that optimizes resource use, reducing costs by 65% while maintaining high model performance.
• (MLOps) To improve manual annotation workflows, I automated data labeling job creation using AWS GroundTruth, reducing the time from 1 hour to 5 minutes accelerating 12x productivity.
Brand Networks LLC | Data Scientist | Remote - Seattle Mar 2019 - Oct 2021
● (NLP/CV) To improve campaign effectiveness, I created a recommendation engine to suggest creative edits based on text and image feature analysis, increasing click performance by 14%.
● (NLP) To improve customer message analysis, I developed an NLP tool leveraging Regex and POS tagging, enabling accurate classification of chat messages and improving productivity by 25%.
Verizon Communications | Data Scientist | Hyderabad, India Jul 2017 - Feb 2019
● (NLP) To improve customer complaint routing, I developed a custom text classification tool, achieving 80% accuracy and reducing misrouting transfer rate by 12%.
● (NLP) To gain chat insights, I developed a PySpark text analytics tool for sentiment, anomalies, and topics, contributing reducing DPMO (defects per million opportunities) by 30%.
● (Recommendation Engine) To increase e-commerce accessory sales, I implemented ALS matrix factorization in PySpark for recommendations, resulting in an increase of annualized sales revenue by 3% ~$14M.
education
Georgia Institute of Technology, Atlanta, US | MS Computer Science
Vellore Institute of Technology (VIT), India | Bachelor of Engineering