• Work on backend services, caching, queueing and DB systems and external service integrations for Rio, a distributed CI/CD platform comprising Rio2 (Jenkins-based) and Codeflows (Kubernetes-based), scaling the platform to handle 1M+ secure builds daily for 11.5k developers across multiple organizations

• Lead multiple high-impact technical initiatives across infrastructure scalability, security, and platform evolution, supporting long-term strategic expansion to non CI & agentic workflows in addition to traditional CI workflows

• Led team transition of next-gen workloads to Apple Cloud KCS (Apple's Kubernetes cluster offering). Collaborated with infrastructure teams as pilot users to shape the offering and enable critical workflows including native amd64, arm64, macOS/CloudOS, and GPU builds on Apple Silicon

• Lead network security hardening in partnership with Apple Cloud Traffic, Identity, and Security teams to implement network access controls and comprehensive audit logging

• Designed and deployed AI agent for automated build failure root cause analysis in CI workflows. Built open-source Python library (logreduce) for intelligent log reduction using MinHash, LSH deduplication, and token-efficient filtering

• Designing and implement CI/CD agent platform backend including context memory systems and reasoning flows that orchestrate build, test, and deployment agents across distributed services

• Built comprehensive observability and metrics collection system capturing customer, operational, and business metrics for data-driven decision-making on feature usage, infrastructure utilization, performance optimization, and capacity forecasting

• Partner with customer teams to deliver solutions that unblock critical workflows and ensure platform reliability through collaboration with SRE and Support teams

• Mentor new and junior engineers on distributed systems, CI/CD architecture, and technical problem-solving

Part of the Platform Engineering Team and worked on the following projects:

1) Personalized Email Services (Campaigns, Ad & Media Analytics)

Delivered highly scalable backend services for personalized FMCG retailer coupon email campaigns to an end user base of ~40 million and track their performance in real time. Developed key features including an A/B Testing framework, campaign management etc

2) Recommendation Engine

Analyzed terabytes of retailer data using Spark to generate coupon offers for users. Also wrote an algorithm for associative rule mining for finding coupons clipped together and basic linear regression for recommending personalized coupons to users.

Designed a data model to store user, offer, and related data in a Graph Database, and built backend services to deliver personalized recommendations to users in real-time, utilizing dynamic configurations in a push model.

Leveraged distributed systems like Redis, Kafka (Streaming), Spark, Janusgraph and Cassandra.

Regularly contributed to engineering tech talks and discussions at Quotient.

Have won 2 hackathons consecutively.

Part of the Platform Engineering Team.

Built and deployed the Image Render Module described below.

Contributed to existing codebase by developing new features and making bug fixes.

Wrote automation bash and python scripts for execution of jobs.

Was part of a 4 person team for developing and optimizing the Adapter Framework.which was responsible for efficiently fetching data from RDBMS, ERP, legacy systems etc. as well as pre processing it. All eQ products are built on top of this framework.

The project aims at creating a system which performs static and dynamic analysis on Android applications to detect suspicious applications.

Built the Android service which monitors real time communication between applications and verifies the inter-process communications against a set of predefined security policies and prevents malicious applications from exploiting transitive permission properties to enable privilege escalation.

Also built an Android Loadable Kernel Module which logs the system calls made by an application during its lifetime.

Coded several machine learning algorithms like K-Nearest Neighbor , K means clustering, J48 tree algorithm to run on the permissions dataset and the system calls dataset gathered to classify applications as malware or benign.