Parhako
Parhako helps students prepare for exams with AI-assisted practice and a bank of 20,000+ questions. More than 2,500 active students use it.
- My work
I led development of the question-processing pipeline: document chunking, vector embeddings, RAG retrieval, structured output parsing, and validation gates. Processing runs asynchronously through SQS queues, with validation retries for generated content.
I built the Python and Redis orchestration layer managing 20+ API keys across five LLM providers. It handles rate limits, exponential backoff, provider failover, and SNS alerts, with background jobs on AWS Lambda. I also handled Docker deployments on EC2 and ECS, S3 asset storage, SES transactional email, and production debugging through CloudWatch metrics and alerts.
- Production
- AWS ECS and EC2, Lambda, SQS, S3, SES, Redis, Docker and CloudWatch.
Warehouse modernization
Warehouse staff needed to use their existing terminal workflows on handheld devices. I rebuilt those screens as a responsive web app that also works offline.
- My work
I rebuilt the legacy worker-terminal interfaces in Angular and TypeScript, with responsive layouts for handheld devices and offline-capable PWA workflows. WebSocket synchronization supplies real-time updates to the interface.
I also built the Java and Spring Boot backend: REST APIs for warehouse operations, WebSocket streaming, and PostgreSQL persistence. The services run in Docker and support logistics teams across multiple sites.
- Built with
- Angular · TypeScript · Java · Spring Boot · WebSockets · PostgreSQL · Docker
Real-time trading system
A trading system that gathers news, scans markets and runs statistical analysis before placing orders through risk checks.
- My work
I architected the system connecting news aggregation, market scanning, and statistical analysis to automated execution workflows. The Python services handle execution, risk checks, circuit breakers, and position tracking, alongside WebSocket connections, PostgreSQL, and n8n workflows.
I containerized the services with Docker and deployed them on AWS ECS, using CloudWatch metrics and alerts to monitor live operations and investigate production issues.
- Built with
- Python · WebSockets · PostgreSQL · n8n · Docker · AWS ECS · CloudWatch
University LMS
A university platform that replaces manual administration of courses, assignments, exams and grades.
- My work
I designed the PostgreSQL schema, Prisma migrations, and REST API layer, including request validation and role-based access control for students, instructors, and administrators.
On the React and Node.js application, I built course registration, assignments, exams, grading, announcements, and dashboards. These workflows replaced manual academic administration; Docker packaged the application for deployment.
- Built with
- React · Node.js · PostgreSQL · Prisma · Docker
Patient management
Patient-management features integrated into an existing healthcare product through embedded Django screens.
- My work
I worked with other engineers on a Django patient-management platform embedded into an existing healthcare SaaS product. The integration used JWT-authenticated iframe sessions, protected routes, and role-based access control.
My work included input validation and PostgreSQL-backed APIs, with patient-management workflows designed to support auditing.
- Built with
- Django · PostgreSQL · REST APIs · JWT
