yedila

Worked on Pearson's AI-powered educational platform, building data-heavy interfaces for automated homework evaluation, feedback and recommendations. Led frontend architecture improvements: extracted business logic from components into hooks and services, increased reusability and optimized rendering under large data volumes. Integrated the frontend with backend APIs, handling asynchronous state, loading and error flows across the application. Implemented accessibility (A11y) as a systematic requirement: semantics, keyboard navigation, contrast and screen reader support.

Built the entire technical infrastructure of a Web3 platform from scratch as the sole engineer: backend, frontend, blockchain integrations, admin panel and deployment. Designed a modular adapter architecture with admin-side configuration instead of coding each partner integration separately, cutting partner onboarding from days to minutes. Implemented multi-step transaction flows (swap → bridge → stake) as persistent state machines with idempotency enforced at the database level, surviving service restarts and preventing duplicate execution. Set up the full delivery pipeline: Docker, Nginx, CI/CD, staging and production environments, backward-compatible migrations and rollback. Result: launched beta on schedule, onboarded 30+ DeFi partners, $10M+ in transaction volume within the first months after launch.

Developed a crypto portfolio management platform across EVM and Cosmos networks, used daily by thousands of users. Diagnosed and fixed a critical production issue where real-time updates silently went stale: connections were never cleaned up on client disconnect, and some were dead client-side but alive server-side. Fixed with active heartbeat detection and proper state cleanup. Reduced transaction latency from 4s to under 1s, cut complex operation time from 90s to 40s, and improved real-time update stability by ~80%. Built a universal wallet adapter and a transaction constructor abstracting the differing rules, fees and finality semantics of each blockchain. Added connection and performance metrics to monitoring, so issues surface on a graph before users report them.

Automated collection and processing of data from heterogeneous external sources, running on schedule with ongoing maintenance. Built a crawler in Python (Scrapy) with a Node.js backend, MongoDB and Redis, handling large-volume batch processing. Result: saved roughly 80% of the content team's time and simplified page publication through Hugo for better SEO.

Led a development team and worked directly with clients from requirements gathering to delivery and support. Launched 10+ projects from idea to production and cut the development cycle by 20–30% through reusable modules. Taught programming for three years in parallel (Web Development, PHP, Python, Django, Node.js, C++) to 100+ students, several of whom now work in the industry.