Founded and built an AI-first trading platform end to end. The product was implemented as a TypeScript Turborepo monorepo with a React trading terminal, a Nest.js backend, Drizzle ORM, PostgreSQL, BullMQ, Redis, and a Next.js documentation site.

Owned features from vague requirement to production deployment across a self-custodial perpetuals DEX aggregator. Used architecture decision records as the standing design record and took responsibility for architecture, implementation, and delivery.

Designed and built an in-house LLM agent as a four-stage cascade: route, research, reason, revise. Routed multiple model providers through OpenRouter behind a single interface, enforced per-stage cost and step budgets, and implemented typed tool calling so the reasoning stage could ground itself in live account and market data while preserving security boundaries.

Contributed to a suite of decentralized trading and finance protocols. Architected a unified DEX abstraction layer behind a single service interface with zero application-layer coupling.

Developed backend infrastructure on AWS using AWS CDK, including APIs for user management, incentives, trading proxies, event tracking, and AI endpoints. Designed and implemented an AI-driven trade suggestion module that produced autonomous perpetual trading recommendations from live market data.

Helped scale the platform to $600M+ in cumulative trading volume and $90K+ in automated referral payouts.

Contributed to a monitoring system for DeFi protocols by collecting and processing on-chain data using Node.js, TypeScript, ClickHouse, and Viem.

Configured Grafana dashboards and custom alerts with notifications to Telegram, Email, and PagerDuty. Integrated APIs and blockchain data via multicall to optimize requests and reduce RPC costs.

Developed a custom CRM system from scratch in a team of two, covering backend and frontend responsibilities. Implemented N-Tier architecture for maintainability and separation of concerns.

Built core modules including JWT-based secure authentication, hierarchical role-based user management, and a Redis-cached export module for performance. Set up CI/CD pipelines with GitHub Actions and configured infrastructure in Azure.

Participated in requirement gathering and business communication directly with the client, defining functional scope through a shared ubiquitous language.

Built and maintained features for hyper- and hybrid-casual games across multiple studios, reaching 10M+ downloads collectively.

Implemented CI/CD pipelines and Git-based version control best practices, reducing release cycle times. Collaborated with cross-functional teams to align features with creative and business requirements.