I build backend systems that survive real-world chaos — retries, failures, race conditions, and unpredictable scale. Focused on reliability, clean architecture, and solving problems that actually show up in production.
Corrdex
Understand Architecture
Corrdex is a codebase understanding system for large repositories. It turns code into structured intelligence about files, functions, dependencies, behaviors, and domains, then uses that structure to answer engineering questions more reliably than search-plus-chat alone.
Most AI coding tools still behave like “chat over a codebase.” They can find files and generate plausible answers, but they often break on follow-ups, ambiguous questions, and architecture-level reasoning. I think real code understanding needs structure first, not just better prompting.
Corrdex does not rely only on live retrieval at answer time. It materializes reusable repository knowledge, routes each question into the right scope, and tries to preserve subject continuity across follow-ups. The goal is to make answers feel more like navigating an architecture system than talking to a generic chatbot.
Most systems look good until real users hit them. Payments fail, APIs break, retries collide, and edge cases pile up. That’s where I work.
I focus on backend systems that deal with reality — concurrency issues, distributed workflows, third-party failures, and scaling without losing control. Clean code matters, but surviving production matters more.
Selected
Works
Fault-tolerant retry system for handling failed transactions.
Handling duplicate payments, retries, and race conditions while ensuring no double refunds or payouts.
Designed idempotent workflows with controlled retries, state tracking, and database-level safeguards. Integrated with unreliable third-party PSP APIs.
Eliminated duplicate transactions and stabilized payment flows under concurrent load.
Distributed backend for managing trip lifecycle with consistent state across services.
Concurrent requests and distributed updates causing inconsistent trip states during high traffic.
Implemented Saga-based workflows to manage multi-step trip operations with eventual consistency and controlled state transitions.
Reliable trip lifecycle management with reduced inconsistencies under concurrent load.
Thread-based chat system powered by Supabase realtime.
Managing multiple chat threads with real-time updates and consistent state across devices.
Built a scalable chat architecture using Supabase realtime subscriptions, optimized thread handling, and local persistence.
Reliable real-time messaging with clean thread management and low latency.
Scalable push notification system with controlled delivery and rate limiting.
High volume of notifications causing rate limits, delivery failures, and inconsistent user experience.
Built queue-based processing with background jobs and rate limiting to ensure reliable and controlled notification delivery.
Improved reliability and delivery success of push notifications under load.
Complex data pipeline syncing large codebases from local IDEs to the cloud.
Transferring massive Abstract Syntax Trees (ASTs) and architecture data from a client's machine to the cloud reliably without timing out or blocking the user.
Engineered a three-tier architecture: a VS Code extension, a local background daemon for parsing, and a highly optimized Cloud Server pipeline (S3 syncing).
Drastically reduced payload sizes and eliminated API gateway timeouts, enabling seamless background syncing of massive codebases.
Work
History
Cabunion
2025 — NowBuilding and scaling backend systems for real-world travel operations, including payments, trip lifecycle management, and reliability under unpredictable production conditions.
Impresiot
2023 Dec - 2024 FebBuilt an event management and ticketing platform end-to-end, including event creation, ticket booking, payments, and user engagement features.
Dev Competitive
Platform
Group-based coding competition platform integrating LeetCode and GitHub. Built to drive consistency through peer competition and real-time progress tracking.
No unified way for developers to track progress, compete with peers, and stay consistent across platforms like LeetCode and GitHub.
Built group-based leaderboards, integrated external data sources, and enabled real-time tracking of coding activity.
Created a competitive environment that encourages consistency and peer-driven motivation among developers.