Contributing to KubeAdmiral, a large-scale multi-cluster Kubernetes management engine used to orchestrate cloud-native applications across cloud-agnostic environments, which manages 210,000 nodes and supports over 100,000 microservices internally.
Resolved a longstanding logging bottleneck in KubeAdmiral by integrating Zerolog with Lumberjack-based log rotation, implementing non-blocking message handling to prevent log blocking during high disk I/O on clusters.
Upgraded KubeAdmiral's internal metrics library and added alerting metrics for dropped logs, enabling on-call engineers to proactively address issues related to the previous log-blocking bottleneck.
Developed a Kubernetes scheduling plugin within the KubeAdmiral scheduling framework to introduce a "sale mode" resource type (e.g., reserved, scheduled, spot, on-demand), enabling capacity-aware workload placement similar to AWS instance types.
Software Development Engineer Intern, Identity Service Team, Autodesk
Maintained and enhanced Autodesk's Authorization API, utilizing the OAuth 2.0 and OpenID Connect standards to authorize requests to Autodesk products for over 20 million monthly active users.
Drove the adoption of Behaviour-Driven Development (BDD) by migrating existing tests from the Go standard library to Ginkgo, collaborating with PMs, SDEs, and QAs to establish a shared understanding of intended software behavior, boosting test efficiency and reliability.
Engineered command-line tools to automate the feature flag archival using custom criteria and markdown reports, eliminating manual reviews and streamlining weekly service maintenance, saving ~2 hours/week.
Worked within an Agile scrum workflow, participating in daily stand-ups, sprint planning, and sprint retrospectives to drive continuous improvement within the team.
HeadCount
Mentored three first-time hackathon participants in developing HeadCount, a real-time occupancy monitoring system.
Integrated YOLO machine learning model with FastAPI for backend and Next.js for frontend development.
Won Best Rookie Hack Award at sunhacks 2024, organized by MLH.
Utilized React, FastAPI, and OpenCV to handle live video processing and people counting.
Full Stack Engineer Intern, Q Team, Home Team Science & Technology Agency
Spearheaded an automated assessment platform for all ten MHA agencies by researching, benchmarking, and integrating state-of-the-art LLMs (incl. LLaMA3.1, Mixtral) to generate and mark questions from training manuals.
Engineered and deployed a full-stack, modular application, designing its architecture with Docker for cloud-based trials, building the UI in React, and securing the FastAPI backend with JWT authentication.
Contributed to the full Software Development Life Cycle (SDLC), engaging with stakeholders to gather requirements, define features, and collaborating with fellow teammates to support the project through to final testing and deployment.
Developed a mobile application that generated food recommendations on campus based on user preference, reviews and timetable data, reducing decision time for users.
Utilized Jest to conduct automated testing, reducing testing time by 50% as compared to manual testing.
Developed a mobile-friendly frontend using React, Sass, Bootstrap and SCSS.
Hosted API calls on the backend via Firebase to ensure consistency and to implement better error handling.