Jun 2025 – Aug 2025 · 2 mos · Gurugram, Haryana, India · Remote

• 1) Built a RAG system using Qdrant (vector DB), BERT embeddings, and locally hosted LLaMA3/Mistral models via Ollama to generate intelligent, context-aware responses.
• 2) Developed an agentic framework integrating Gmail, Google Calendar, Tasks, and Slack for unified natural language access to user data and multi-step workflow execution. Powered by GPT-4-mini, a custom MCP server with streamable HTTP transport, and built using Python, FastAPI, OpenAI’s Agents SDK, and Postgres for chat persistence.
• 3) Engineered a cross-platform system app (Swift + Electron) for live audio transcription and ingestion into the RAG pipeline, enabling real-time capture and retrieval of personal knowledge as part of the agentic tool suite.

Aug 2024 – May 2025 · 9 mos · Gurugram, Haryana, India · On-site

• 1) Engineered a recon microservice backend module for dynamic document generation (invoices, credit notes, summaries) using Spring Boot, Thymeleaf for templating, and applying SOLID principles plus design patterns for a clean, extensible service layer.
• 2) Developed a master-data microservice with Spring Boot, Spring Data JPA and MySQL REST APIs for workshop and pricing management, and offloaded memory-intensive bulk operations to GCP Cloud Functions.
• 3) Implemented a proof-of-concept multi-tenant architecture (same DB, multi-schema) with tenant resolution and data isolation at the backend, enabling scalable onboarding of multiple clients.
• 4) Ensured backend reliability with 90%+ test coverage using Groovy (Spock) unit and integration tests, validating service endpoints, data access, and business logic.

Nov 2021 – Aug 2024 · 2 yrs 9 mos · Noida, Uttar Pradesh, India · Hybrid

• 1. Architected Scalable Ad-Serving System:
• Designed and implemented a scalable and robust Ads Recommendation Server which served personalized Ad banners to Paytm's traffic.
• Developed Reactive Kafka consumers to consume the relevant user's data of clicks and impressions on Ads and used caffeine cache to store the computed Ad ranks which can cater to throughput of upto 800 QPS by a single Kubernetes pod with a p99 latency under 2ms.
• Improved user's engagement on Paytm App and a notable uplift in CTR for Ad campaigns.
• 2. Optimized Redis Data Insertion and Tuning:
• Developed a feature for the insertion of ~600 GB of data from hundreds of CSV files into Redis within 3-4 hours using pipe/bulk operation feature of Redis.
• Used state design pattern making the insertion feature scalable and easy to maintain and debug.
• Tuned Redis cluster to reduce infrastructure costs by $22k/month, ensuring optimal performance and resource utilization.
• 3. High-Throughput User Data Micro-service:
• Designed and implemented a micro-service for state management of user-related data, coupled with a Redis lookup mechanism using spring reactive.
• Achieved a throughput of 20k (QPS) with a p99 latency under 10ms.
• Enabled immediate updates in audience targeting for Ad campaigns, significantly reducing the time from a day to real-time effectiveness.
• 4. Enhancements to Existing Micro-services:
• Developed various features within existing micro-services, including RESTful APIs, internal caching, and AWS Lambda functions.
• Implemented robust monitoring mechanisms and alerts to ensure system stability and performance.
• Demonstrated proficiency in writing Unit Test Cases (UTCs) for reliable and maintainable code.

Sep 2020 – Oct 2021 · 1 yr 1 mo

• Worked on Finastra Payments Product, GPP (Global PayPlus), a Monolithic Application.
• Projects:
• 1) Designed and Implemented the Integration of Java Server, a low-latency, real-time and secure
• Maven Application with GPP (Global PayPlus), a Monolithic Application, following n-tier architecture
• for Customers around the world.
• 2) Collaborating with agile development team, designed a POC i.e. a C++ application and a multi- threaded Java Server communication service using Unix Domain Socket for AIX machine.
• 3) Active role in defect fixing, extensive code debugging, unit testing and handling ad-hoc requests
• from GPP clients while providing on-time and on-budget deliveries.

Jan 2020 – Aug 2020 · 7 mos

• 1) Built an Automation Tool using selenium and chrome driver to automate the fetching of payments
• data from swift's portal and store the results in excel sheet using poi jar.
• 2) Collected, retrieved and debugged data for close analysis of product infrastructure and payments.

Jul 2019 – Nov 2019 · 4 mos · Pune, Maharashtra, India

Nov 2018 – Dec 2019 · 1 yr 1 mo · Vishwakarma Institute of Technology

• Part of Student Satellite Team comprising of members from interdisciplinary studies.
• The raison d'être of my team was handling the orientation of the Satellite according to the requirements of the payload by building various mathematical models and simultaneously testing these models on our satellite.

Feb 2018 – Oct 2018 · 8 mos · Vishwakarma Institute of Technology