1. Migrated the loan collections account processing from Cron to Kafka, reducing processing time by 60% and ensuring data availability 3 hours before business hours for over 9M active accounts.2. Reduced payment failures significantly by optimising the database, moved the payments module to Kubernetes (K8s), and eliminated 80-100 daily lost requests due to unscalable code and DB failures, ensuring real-time processing with no requests lost.3. Integrated Paytm Golden Gate app with Paytm Creditmate, leveraging Maxwell and Kafka for real-time updates on daily loan case allocations, handling approximately 900,000 cases (30k cases per allocation) across 20-30 campaigns daily. Implemented a system to consume 500 records at a time to match Golden Gate's rate limiting, used DLQ (delayed Kafka) to handle failure cases, and employed the Factory design pattern to achieve this.4. Revamped the bucketing and resolution calculation of loan cases using Spark and EMR, optimizing costs with Transient clusters and r6g instances for in-memory processing. Utilized EMR's Primary, Core, and Task nodes to process 20M cases daily for major Paytm products like MCA, PL, and PP, ensuring data readiness before business hours.5. Onboarded the EDC and Soundbox lender by automating account creation and payments using API and Kafka, used Redis to handle rate limiting for fetching the latest dues, and managed collections for 1.5 million accounts monthly, ensuring data consistency between both systems. 6. Experienced with Shedlock crons in Java, using AWS CloudWatch to trigger Lambda functions for various operations, and working with AWS S3 for data fetching and storage.7. As part of on-call duties, consistently provided RCA for 4xx and 5xx major outages and bugs reported by end users or the business team, investigated API latency issues, and worked diligently to enhance Operational Excellence.8. Prometheus metrics, Grafana dashboards, and alerts were set up for new feature deployments.

1. Worked on improving the code coverage of repositories using JUnit.2. Improved repositories by fixing Sonar issues using ESLint.3. Integrated Vault to ensure the security of application secrets.4. Implemented JWT-based authorization in cross-team APIs.

1. Developed a system that compares the results and the training time taken by the classical techniques of machine learning and the quantum machine learning.2. From the behaviour of GANs to mimic real data, an effort was made todecrpyt the lightweight ciphers.

1. Developed a deep-learning solution for the cryptanalysis of lightweight block ciphers.2. Worked with team members to create a system to generate 1 lakh random ciphertext-plaintext pairs of the lightweight cipher Present.3. With proper dedication of teams, a deep learning model using Tensorflow and Keras was build and tested on Google Colaboratory.4. Finally, we were able to predict (with 50 % probability) each bit of plaintext directly from ciphertext using the above model.