projects

Built Core AI Coding Agent, Editor and Agentic Chat experience for VSCode and JetBrains — used by 2K+ engineers daily, with ~95% shared cross-platform code.

• Architected and delivered MCP integration for the Coding Agent, building frontend and backend infrastructure for server onboarding, tool discovery, permission management, and agent-driven execution of external tools.
• Led end-to-end development of the AI-native Editor and Agentic Chat Experience for the agent, including editor, tool calls, streaming, attachments, @mentions, slash commands, image paste/upload, drag-and-drop, and keyboard-first interactions.
• Improved AI coding agent response quality by introducing compile-time IDE diagnostics as an iterative feedback loop and designing XML-based high-signal context for prompts.
• Built context providers powering the AI agent workflow: @Files & Folders, @Commit, @Branch, @Rules, @Docs, @Linter Errors, @Recent Changes, and @Past Chats.
• Contributed to a cross-IDE platform architecture enabling ~95% shared code between VS Code and JetBrains, significantly reducing duplication and increasing shipping velocity.
• Built extension and agent settings infrastructure end-to-end, including persistence, real-time synchronization, permissions/approval systems, and extensibility surfaces.

ML-driven system predicting which tests should run per code change in local and CI/CD pipelines, using changed files and historical failure signals.

• Shipped infra which used ML models to predict which Unit, Integration, and VR tests should execute for a given code change across local and CI pipelines.
• Reduced test execution volume and improved developer feedback loops, resulting in faster build times, improved productivity, and lower infrastructure costs.
• Drove a ~99% improvement in p95 prediction API latency (60s → 500ms) by implementing Tecton feature store integration and Redis caching alongside FastAPI migration.
• Integrated Tecton as an online/offline feature store to ensure training-serving consistency, enable low-latency feature retrieval, and reduce feature drift in real-time predictions.
• Contributed to MLOps infrastructure automation across the ML lifecycle: data ingestion, validation, labeling/splitting, training, retraining, testing, deployment, and serving.

AI-powered image editing and generation platform with background removal, object replacement, and AI avatar generation. Reached 700K+ users.

• Led a team of 5 engineers to build the frontend platform, owning complete architecture, AI feature integration, performance optimization, SEO, CI/CD pipelines, and production deployments
• Built AI-powered image editing capabilities including Background Remover, Object Replacer, and AI Avatar generation workflows.
• Integrated Stripe for secure payments, enabling subscription management, checkout flows, and seamless transaction handling across the platform.
• Implemented Firebase One-Tap Google Authentication and optimized image editing workflows via async processing, state management improvements, and reduced UI latency.

Cloud-based storage and PDF toolkit platform with real-time file conversion, custom PDF viewer, and document management. Reached 500K+ users.

• Led a team of 3 engineers and built the frontend platform for Cloud Storage and Document Conversion
• Built Redux based Architecture for the cloud storage platform, powering file uploads/downloads, CRUD operations, nested folder trees, and storage management workflows.
• Built real-time file upload and conversion workflows using Socket APIs.
• Developed a custom PDF viewer enabling in-browser document viewing and interaction.
• Integrated CAPTCHA-based bot protection and reduced abuse and malicious traffic and improved security