San Jose, California, United States
- Founding engineer on a 10-person team building a mobile-first non-custodial wallet platform from ground up - Owned and designed Legend’s async task-processing framework – exactly-once delivery, stateless recovery, and Pulumi-driven auto-generation of resources on GCP (topics and subscriptions) which lets engineers to easily build new processors on top of the framework; became the backbone for all Legend’s background task processors - Developed QuarkWallet, a smart contract wallet implementing account abstraction primitives (transaction batching, gasless transaction, conditional transaction) - Developed Tradewind, an intent-based routing engine converting complex user intents into batched execution flows across multiple chains - Designed Quark Core Scripts, a library of composable on-chain execution primitives (PayCall, QuoteCall, and others) that let QuarkWallet bundle multi-step actions into single atomic trans- actions - Led and developed a unified Elixir notification platform powering real-time iOS push alerts and MJML-templated transactional emails - Integrated Defi protocols Compound, Aave, Morpho, and Circle CCTP to support lending, yield generation, and cross-chain USDC transfers
- Contributed to the core development of a decentralized lending platform (peaked ~$8B TVL) on blockchain technology. - Developed a robust Elixir backend system for Compound Treasury, achieving $27M in AUM. - Created a TypeScript backend serverless API for efficient on-chain data parsing and transaction readability.
- Maintained 99.99%+ uptime for Gmail backend infrastructure as a primary 24/7 on-call engineer - Engineered alerting systems, and deployment pipelines governing safe production rollouts - Supported infrastructure reliability initiatives and incident response across distributed systems
- Bing Hotel Ads Infrastructure - Led 2 junior engineers working on Bing Hotel Ads - Engineered scalable Azure processors ingesting and serving real-time ad data to Bing.com - Developed reliable high-throughput offline big data workflows using Microsoft Cosmos
- Designed and Developed Orchestrator project in Node.js (in production for IBM Support Portal) from the ground up with senior developers that leverages IBM Watson's cognitive power in Test-driven Development(TDD) approach - Developed automation script for training Retrieve and Ranker and building Collections on ETL server which reduced process time from days to hours - Composed several patent disclosures in the area of cognitive and social computing with fellow interns, one of which is being drafted by a patent attorney. - Wrote View function on Cloudant Database for improving the efficiency of retrieving stored documents from it - Experimented the performance of the Ranker by modifying the collection contents in Solr Retrieve Server Note: Orchestrator is serving all client's queries by leveraging several Watson's capabilities like Retrieve and Rank and Conversation. Orchestrator is now a key component of Future for Support at IBM and handles traffic across all interface for client (web, chat, phone, home agent, etc.)