Jan 2024 – May 2024 · 4 mos

Jan 2023 – Jan 2024 · 1 yr · New York City Metropolitan Area · Remote

• Open source engineer, worked on Flyte and Union Cloud Console, a performant workflow orchestration platform that seamlessly unifies data, ML and analytic stacks written in Go, Python and React.
• Implemented gRPC services for supporting complex real-time workflow visualizations and executions over DAG
• Implemented custom parsers and plugins for quick-compilation of run-time inputs built on top of JSON schema spec.
• Designed and implemented a highly optimized KDTree based algorithm for Distribution and Alignment snapping of workflow elements
• Programmed observability metrics for Union Cloud platform for real-time monitoring of logs and important metrics such as GPU, CPU utilization, live streaming of workflow execution logs
• Revamped Union Cloud Platform design using Atomic Design

Dec 2022 – May 2024 · 1 yr 5 mos

• Spearheaded development at Antimetal as Founding Fullstack Engineer, architecting solutions using Next.js(React), TypeScript, FastAPI (Python), and AWS infrastructure including Lambda, ECS, and DynamoDB
• Led Frontend Architecture as sole developer for the Antimetal Save Platform, implementing real-time data visualization, state management with Zustand
• Engineered a sophisticated recommendation engine utilizing statistical modelling and heuristic approaches to analyze AWS resource utilization patterns, achieving 25-30% cost reduction for beta customers through rightsizing and resource termination strategies
• Designed and developed distributed microservices in Go and Node.js with message queuing (ARQ), implementing custom ETL pipelines for cost forecasting with statistical modeling, metrics collection for comprehensive observability, and anomaly detection using deviation analysis algorithms

Dec 2021 – Dec 2022 · 1 yr · New York City Metropolitan Area · Remote

• Worked on Synthesia Studio, a performant in-browser video editor written in React.js with serverless Flask API with PostgresQL and MongoDB.
• Implemented support for complex real-time collaboration and presence detection in Synthesia Studio in React using Conflict-free Replicated Data Types (CRDT)
• Implemented custom plugins and mark wrappers using Tiptap and Prose-mirror to restructure the speech input and add support for Synthesia specific nodes to annote events in videos like gestures, micro-gestures, timeline markers, etc.
• Designed and implemented a highly optimized KDTree based algorithm for Distribution and Alignment snapping of canvas elements for Synthesia Studio
• Extended Synthesia's internal SSML parser with support for customised tags for timing video animation with voice script
• Revamped Synthesia STUDIO Editor design using Atomic Design

May 2021 – Aug 2021 · 3 mos

• Remodelled existing code on NVM to rollout support for Windows-based systems
• Refactored npm-lockfiles to integrate support for Arborist
• Implemented a CLI tool using Go to monitor Github pull requests for checks using polling and notify user when ready to merge

Feb 2021 – Apr 2021 · 2 mos

• Created WebVR samples that interface with WebXR MediaLayers to achieve fluid resizing & motion in 3D space
• Extended Three.js adding support for controller & hand interactions
• Impact Illustrated improved performance, legibility with 57% more battery-life & 17% reduced latency

Jan 2021 – Dec 2021 · 11 mos

• Developed several internal visualization tools for tracking machine learning pipelines and models using React.js and D3.js
• Implemented end-to-end Alan STUDIO, a React.js based browser environment for allowing users to train Alan’s AI bot right within their own browser
• Built a synthetic dataset sampler to train Alan’s entity-recogniser, improving accuracy by 4.6%
• Added multilingual support for Alan’s entity-recogniser
• Dockerized existing scripts for running model training on cloud
• Worked on Node.js and Apollo GraphQL to develop a performant GraphQL backend server
• Integrated a robust queue system using BullMQ for processing heavy background jobs
• Impact Improved server query response time by 72% by restructing API and optimizing database queries
• Led efforts for Alan Studio, a cloud-based service that let’s users train and test Alan’s capabilities in less than 30 seconds
• Reduced reporting API endpoint response times from several minutes to fewer than 7 seconds by rewriting the API using Nestjs, SQL query optimization and caching
• Built a group of microservices to automate infrastructure management for Alan Platform using Nestjs, Caddy proxy, Terraform, Google Cloudrun, Google PubSub

Jan 2020 – Sep 2020 · 8 mos

• Created a Jest-preset to bootstrap test environment for Mediawiki tests
• Added test video recording with Jest using FFmpeg & Xvfb server
• Upgraded wdio-video-reporter npm package to add support for WebdriverIO v7

Aug 2018 – Dec 2018 · 4 mos

• Part of Address Intelligence Team, Last Mile Technologies. Worked on Global Address Model’s Sanity module
• Created automated test data generator which is triggered in case of Sanity failure of nightly builds of GAM,
• bringing down the failure debugging times from 5 hours to < 15 mins.

May 2018 – Aug 2018 · 3 mos

• Developed a Javascript SDK to mock API calls from frontend using Service Worker API.
• Designed a typescript codegen for automatically generating SDK API mocks based on a schema file.
• Setup a complete test system using Puppeteer and Jest that consisted of 5 modules which validated the APIs for spec compliance

Dec 2017 – Mar 2018 · 3 mos

• Refactored their CI/CD automation bot & made it configurable to multiple Github repositories
• Trained a probabilistic model on Falkonry's own historical logs to predict network outages & cause of outages
• Impact Implemented logic to efficiently parse historical logs saving 38% computation time