Experience

Staff Machine Learning Engineer

PayPal, San Jose, CA | Jul ‘23 - Present

Generative AI Adoption: Led the adoption of Generative AI at PayPal with a focus on safety and responsible AI, streamlining processes.
Developer Productivity: Boosted developer productivity by 30% through successful PoCs in software and ML development, shaping strategic decisions.
LLM Lifecycle: Developed an end-to-end lifecycle for Open Source Coding LLMs, inclduing Finetuning, RAG, Agentic Workflows, Eval and Deployment

Standard Cognition, San Francisco, CA | Nov ‘19 - May ‘23

Autonomous Checkout Scaling: Expanded autonomous checkout from 2 to 40+ stores with increased complexity, foot traffic, and hardware diversity, leveraging ML models like Human Pose Estimation.
Next-Gen Hardware Design: Delivered edge-based perception hardware with a 6-month roadmap, improving margins by $1M/store/year.
Retail Analytics Innovation: Established verticals in Human Trajectory Analytics and Visual Re-Identification, boosting profit margins and user retention.
Automatic Labeling System: Developed an auto-labeling system for tail distribution detection, increasing deployments 10x and cutting development costs by 85%.
ML Operations Efficiency: Designed systems for automated ML model workflows, reducing operational costs by 90% and improving accuracy by 25%.
Vertex AI Platform: Built an ML inference platform from scratch, serving 5 internal ML teams.
Leadership and Team Management: Led and mentored a global team of 7, fostering team culture, career growth, and cross-functional collaboration.
Strategic Roadmapping: Defined team strategy, roadmap, and success metrics, translating them into actionable engineering goals.

NIO, San Jose, CA | March ‘18 - Nov ‘19

Edge Compute Design: Led the design and selection of edge compute for NIO’s fully autonomous (L4) platform, foundational to the Adam supercomputer powering 100k+ vehicles/year.
DNN Optimization: Accelerated ML models up to 5x for edge deployment in L4 autonomous vehicles with minimal accuracy loss (<1%).
Autonomous Driving Algorithms: Designed optimized DNNs for lidar-based scene understanding, including a two-stage object detection algorithm with multimodal inputs, achieving compressed and accelerated performance.
Custom Inference Library: Built a low-level library for edge inference, enabling deep learning model support across various hardware accelerators.
Team Mentorship: Guided senior engineers and interns in deploying multiple DNN models concurrently on embedded platforms.

Otsuka Digital Health, Princeton, NJ | July ‘17 - March ‘18

Medical Claims Prediction: Enhanced short-term cost prediction for medical claims by 90% using Deep Learning (LSTM) to prioritize patient care.
Advanced ML Solutions: Developed and implemented predictive models and innovative methods to analyze and visualize medical claims data.

Rutgers University, New Brunswick, NJ | July ‘16- June ‘17

Face Clustering: Developed a novel algorithm for face clustering based on multiple facial attributes.
Face Recognition: Developed and published a novel face recognition algorithm for aggregating visual features based on clustering in a multi-shot video-to-gallery template retrieval problem in an unconstrained environment.
Photo-Sketch Generation: Investigated the role of face data and attribute bias in automated photo-sketch generation.

Indian Institute of Technology, Delhi, India | Dec ‘11 - July ‘15

Speaker Verification: Developed and published a robust speaker verification algorithm invariant to noise and multi-channel input using GFCC, MFCC and i-vectors.
Eye Movement Tracking: Developed and published novel decision tree based method for error analysis for eye movement tracking for biometrics.
Iris Dataset: Independently led data collection, curation, labeling and management of human irises datasets from 50+ users in a period of one month.