Led the design for building Experimentation Platform framework for alignment techniques such as instruction tuning, reinforcement learning from human feedback (RLHF), and parameter efficient fine-tuning such as p-tuning, adaptors, LoRA, and so on to improve use cases.

Led the design and implementation of a new micro-services architecture, resulting in a 30% improvement in system performance.

Mentored junior engineers, helping them enhance their technical skills and contribute effectively to the team.

Collaborated with product managers to define and prioritize features, ensuring alignment with business goals.

Implemented best practices for code quality and testing, leading to a 20% reduction in bug reports.

Collaborated with 13 Senior engineers SME group for Amazon – we are responsible for creating Amazon SDE Online Assessment Coding Questions to hire SDEs across Amazon. We architected about 50+ Amazon SDE Coding questions which has been used to evaluate more than 20K SDE 1s & 2s globally by the recruiting engine

Built a Computer Vision model to track the dwelling pallets. Improved Precision to 97.4% and recall to 83%

Designed the next-gen platform for data extraction and pipeline building. We improved the data interoperability between Redshift, MySQL, S3, RDS and Emailer. Designed the platform with Apache Airflow 2.0, AppSync, MWAA to feed metadata, DynamoDB, CDN and React

Designed Time-to-Live mechanism that bridges the data in-flight between Amazon DynamoDB and Kinesis data streams.Built the record processor to compile parallell on the checkpoints of the processed records of all the shards that it manages and reduced memory overhead by 17%.

Spearheaded a team of 4 SDEs + 2 DEs, to build a system similar to Git for archival of huge data file and to pre-process 14TB data based on the authentication levels. Tech stack involves AWS streams, EMR, ECS, Kubernetes, DynamoDB, Redis, S3 and REST

As part of Operational Excellence, I authored a document on Effective Design Document Writing which is being used by more than 33+ project designs

Built a Continuous Deployment Pipeline mechanism Edge Deployer using MQTT topics to enable wave deployments and to perform load balancing effectively. The sensor fusion mechanism was later converted into a library construct which is being used by more than 100+ teams across Amazon

Identified a real-time problem to address safety hazards at 350 Amazon FCs. Built a deep-learning model to eliminate the potential human safety hazards, reduced the manual touchpoints by 87%