Dorna Gyawali

I’m an ICT student focused on building production-minded software—especially systems that are low-latency, scalable, and measurable. I care about clean architecture, reliable pipelines, and user-facing performance. In my Quark project, I built an AI RAG system with persistent memory. I designed a Redis + BullMQ worker pipeline to keep APIs consistently low-latency under concurrent load, implemented hybrid memory (Redis + vector store), and improved retrieval accuracy using an embedding + reranking pipeline. I also build real-time and distributed systems. With SDR-9, I created a browser-based P2P file transfer system (1GB+) using chunked streaming and Socket.io signaling, while offloading background email and notifications to Redis worker queues. With PX47, I developed a distributed media processing pipeline that uses pre-signed S3 uploads and FFmpeg-based multi-stage transformations for scalable async transcoding and waveform/metadata generation. Beyond projects, I contribute to GreedyBear (IntelOwl ecosystem), working on threat intelligence IOC analysis using Python/Django tooling and ML-supporting observability. I’m excited to apply this same engineering mindset to teams that value impact, reliability, and continuous improvement.