Built large-scale distributed systems and machine learning platform for handling 100s of terabytes of data and forming datasets for machine learning model training.Architected backend with 6 microservices on AWS to power a SaaS product for video and image data analytics using deep learning, machine learning & computer vision algorithms for critical decision-making.Built a GPU based scalable, fault-tolerant, distributed data processing service on AWS, using Kubernetes, Kafka & Redis, capable of handling more than 1.7 terabytes of video data per day.Orchestrated pipelines for end-to-end machine learning model development including data extraction, training, tuning and evaluation using Python, PyTorch, MLFlow, Apache Spark & Apache Airflow; used for more than 700 model training cycles.Optimized deep learning inference pipelines to improve throughput by > 2x by refactoring from Python and PyTorch to multi-threaded C/C++ and TensorRT.Developed & maintained a highly-available REST API for serving machine learning inference using Python Flask framework, PostgreSQL database, AWS API Gateway & AWS Lambda.Designed a schema for logging system events, performance and behavior in Elasticsearch and developed Apache Spark applications to analyze those logs for insights into performance.Collaborated with 9 engineers to build internal libraries for platform-wide utilities (such as auth, logging, etc.) that helped reduce 100s of lines of boilerplate code across all microservices.Maintained production codebases by improving unit tests, conducting code reviews and enhancing CI/CD processes.Experimented and developed proof of concept prototypes with various languages and technologies such as Go (Golang), Java, Apache Cassandra, RabbitMQ, gRPC, protocol buffers, etc to improve product platform's system design.

Developed high-reliability, real-time military communication software using C & C++ on an ARM platform using FreeRTOS and Linux to meet stringent latency requirements.Utilized multi-threaded functional programming paradigm to build large-scale distributed systems & event-driven systems.Automated the testing of complex systems by building integration test suites using Python.Built thread-safe C & C++ API libraries for FPGA hardware modules and serial, Ethernet communication interfaces to improve cross-functional teams’ development speed on embedded platform.

Developed Perl scripts for data extraction, sense-making and analytics from millions of lines of logs for product engineers to quickly derive insights from FPGA silicon performance.Built a web app using PHP and MySQL for storing status and test results of silicon chips which helped engineers run tests and track defects more efficiently, thereby increasing chip manufacturing yield.