Kelvin Vora

Developed Python automation scripts and backend tooling for an AI evaluation and grading platform, reducing manual orchestration by 40% and enabling scalable assessment workflows across MCP-based applications. Architected and deployed 7+ production web applications using Docker and Google Cloud, cutting environment setup time by 30% and improving deployment consistency and scalability. Built and maintained an internal task management platform using React, Node.js, REST APIs, Docker, and GitHub Actions. Implemented model API integrations for evaluation workflows including token usage management, request-limit handling, automated grading logic, and model performance evaluation. Created multi-task creation workflows, role-based access controls, evaluation dashboards, and client-facing reporting. Containerized open-source applications and integrated them into automated CI/CD pipelines; collaborated on architecture decisions, code reviews, debugging, testing, and production deployment.

Migrated application state management from Context API to Redux within a React.js + RxJS codebase, improving state consistency and reducing reported bugs by 25%. Developed reusable React components and frontend functionality to improve maintainability, scalability, and cross-browser compatibility. Implemented GraphQL resolvers using Node.js, Express.js, GraphQL, MongoDB, and Google Cloud Storage. Designed and optimized MongoDB schemas to support business logic while balancing normalization, query performance, and evolving requirements. Created and documented backend API specifications using Swagger/OpenAPI to establish clear API contracts and improve developer/QA workflows. Collaborated with developers using GitHub and Agile practices to deliver production features.

Built a social media mobile app (Stikbook) using React Native, Expo Router, and Redux, delivering UI/UX work in Figma and implementing authentication, search, and settings within 6 weeks in an Agile environment while meeting accessibility standards. Collaborated with cross-functional teams to align frontend and backend integration; used Postman for API testing and integration and presented features to stakeholders prior to release. Trained and fine-tuned a video tagging engine model to improve image classification accuracy using Python and TensorFlow, experimenting with multiple datasets and open-source models.