• Conceptualized, developed, and launched an Android app replicating notification LED functionality, amassing over 5 million downloads and user base of 200K+ weekly users. Recognized by XDA and Android Authority for innovation and impact.• Led the full software development life-cycle, managing design, UI/UX and testing. Implemented Bitrise for CI/CD, ensuring seamless updates and quality assurance.

• Engaged in developing machine learning models to forecast fluctuations in commodity prices based on trade features. Additionally, focused on enhancing model performance through the implementation of EDCR technique, a neuro-symbolic approach.• Conducted research work on identification of spoofed AIS data in shipping vessels using attention based deep learning models built using PyTorch.• Managed and enhanced 10TBs of AIS data (production grade) utilizing Hadoop, Spark, and SLURM for efficient processing and analysis.

• Payments: Developed and maintained Java Springboot microservices for Expedia's payment workflow driving $270M per day in revenue.• Fine Grained Authorization: Implemented fine-grained authorization using Google's Zanzibar on a gRPC microservice, resulting in a reduction in unauthorized access to user data.• Dashboards: Led the effort to set up monitoring dashboards and alerts for our team's microservices using Datadog and Splunk, which reduced the time to detect and resolve incidents by 60%.• Real‐time Notifications: Optimized real-time notifications system powering promotional and booking communications by leveraging Apache Kafka, Apache Storm, and Redis cache, resulting in increased user engagement and customer satisfaction.• Batch processing: Rebuilt batch jobs consuming terabytes of data in Apache Spark to drastically reduce execution time by six-folds and decrease execution cost by 50%.• ETL Jobs: Leveraged technologies such as AWS Elastic MapReduce, AWS Datapipeline, Qubole, AWS Kinesis to run ETL processes that drove business analytics tool.

- Virtual Agent: Delivered an end-to-end virtual agent system using Python, FastAPI, and NLP techniques, which achieved a 95% accuracy rate in recognizing customer queries.- FastAPI: Used FastAPI's asynchronous architecture to create scalable backend services for the virtual agent.- Natural language processing: Experimented with SVM, Naive Bayes, and Logistic Regression for text classification and achieved a 90% accuracy rate with SVM.