Aniket Biswas

Understanding problems. Building useful software.

I’m a Software Engineer II at Microsoft, currently on the OneDrive and SharePoint team. My work has ranged from product features and native Android integrations to notification systems and applied AI.

I’m building the AI Harness, a contract-driven framework for spec-validated, LLM-assisted engineering. I’ve used it to lead development of an app-revamp feature in an existing codebase.

Résumé
Portrait of Aniket Biswas

Technologies

  • JavaScript/TypeScript
  • React/React Native
  • Node.js/Go
  • GraphQL/REST APIs
  • Python/TensorFlow/Keras
  • MCP & LLM Integration
  • OpenAI & Anthropic APIs
  • Azure AI Foundry

Experience

  1. March 2023 - Present

    Hyderabad, India

    Software Engineer IIMicrosoft

    Current team: OneDrive and SharePoint

    • Built a native Android WebView module for React Native with disk-based caching and request interception, reducing load time from 12.9 seconds to 5.5 seconds
    • Led the SharePoint News API migration, including architecture and UX changes, while mentoring 2 engineers
    • Shipped React Native features to 3M+ users while maintaining 99.9% reliability
    • Built a config-driven component framework that made feature development and partner onboarding 40% faster
    • Authored the AI Harness, a contract-driven framework for spec-validated LLM-assisted engineering
    • Led development of a brownfield app-revamp feature using the AI Harness, achieving an approximately 5x improvement in development productivity

  2. July 2020 - March 2023

    Hyderabad, India

    Software EngineerMicrosoft

    • Designed a low-latency notification system with clear API contracts and consistent UI state across partner apps in Microsoft Teams
    • Implemented data scoping for 5+ enterprise scenarios sharing one JavaScript runtime
    • Developed a link interception module that routed 2M+ users to the right in-app experiences with 99.9% production reliability
    • Built in-app diagnostics that reduced debugging time by 60+ hours per month
Earlier experience 3 roles

  1. May 2019 - July 2019

    Bengaluru, India

    Research Intern (Deep Learning)Samsung Research Institute

    • Trained a TensorFlow and Keras model to replace manual calibration for wearable controllers using live sensor data
    • Designed a specialized RNN and custom loss function that reduced error by 75%

  2. May 2018 - June 2018

    Bengaluru, India

    Data Science InternPrakshep

    • Used Sentinel-2 imagery, Python, and OpenCV to classify water bodies and agricultural land
    • Built Z-score models over spectral indices to flag harvest windows and outlier plots for precision farming

  3. May 2017 - August 2017

    Bengaluru, India

    Data Science InternFarebond

    • Built a flight-fare-lock pricing model using flight volatility
    • Automated reports with R and R Markdown, saving 60+ hours of manual data entry per month

Projects

  1. macOS Interface Playground interface

    Project 01

    macOS Interface Playground

    A macOS-inspired interface built with React and a Go backend. It explores server-driven UI with dock navigation, window management, and desktop interactions.

    • React
    • Go
    • Server-Driven UI
    • TypeScript
    • REST APIs
    • macOS Design System
    • Responsive Design
  2. Obsidian MCP Server interface

    Project 02

    Obsidian MCP Server

    A TypeScript and Node.js MCP server that connects AI clients with Obsidian through its REST API, supporting note search, creation, retrieval, and organization.

    • TypeScript
    • Node.js
    • Model Context Protocol
    • REST API
    • LLM Integration
    View source GitHub

  3. Project 03

    EthPhoto

    A decentralized application for geo-tagged media sharing using Ethereum smart contracts and IPFS, with an Embark-based JavaScript and jQuery interface.

    • Ethereum
    • IPFS
    • Solidity
    • JavaScript
    • Embark
    • jQuery
    View source GitHub
  4. Personal Portfolio Website interface

    Project 04

    Personal Portfolio Website

    A portfolio site built with Gatsby, React, GraphQL, and Styled Components. Experience and project content is maintained in Markdown; the Writing section loads the Medium feed on page visits.

    • React
    • JavaScript
    • GraphQL
    • Styled Components
    • Gatsby
    • Responsive Design
    View source GitHub
  5. LoanLens interface

    Project 05

    LoanLens

    A browser-based planner for new and existing EMI loans in rupees. Compare monthly payments, prepayments, interest costs, and payoff timelines for reducing-balance loans. Explore repayment schedules and export scenarios, with optional on-device saving and no sign-in.

    • JavaScript
    • Browser Modules
    • Local Storage
    Open demo

Writing

All articles on Medium

Certifications

  1. Coursework
    Anthropic

    Model Context Protocol: Advanced Implementation

    MCP server design and implementation patterns for AI integrations.

    Date
  2. Coursework
    Anthropic

    Introduction to Model Context Protocol

    Fundamentals of Model Context Protocol (MCP) for AI agent integrations.

    Date
  3. Coursework
    Microsoft (via Skillup AI IDC)

    AI Readiness: Intermediate Proficiency

    AI concepts, tools, and practical applications.

    Date

Say hello

Contact

Have a question or want to say hello? You can reach me by email.

Elsewhere online

GitHubLinkedInLeetCode