Associate Software Developer @ Matrix Comsec | Spring Boot · Kafka · FastAPI | Distributed Systems | LeetCode Top 1%

github

Vyom Shah  • Peerlist

Competitive programming platforms face two major challenges: scalability during contests and the tedious process of creating test cases. We built CodeSphere, a full-stack platform, to tackle these issues head-on. To handle high submission loads, I designed and implemented a distributed master-slave architecture that processes code executions in parallel, ensuring high scalability and significantly cutting submission wait times. The platform features a secure contest mode with tab-switch detection and auto-submission, reducing cheating. One of the most impactful features is an AI-powered test case generator I developed using the Gemini API. This tool slashed the manual effort for contest creators. The entire backend is built on a suite of modular REST APIs for maintainable and decoupled management of problems, contests, and scoring

CodeSphere - Real-Time Competitive Coding Pla

How can you transfer files directly and securely between two devices without relying on a central server or cloud storage? We built this P2P File Transfer application to answer that question. The project is a Blazor web app that uses a server-less architecture orchestrated with WebRTC for the direct peer-to-peer connection and SignalR for the initial signaling. This approach completely bypasses the need for cloud intermediaries or user authentication. By optimizing the signaling process in ASP.NET Core, I was able to cut the connection setup time by 40%. Performance was further streamlined by using the Stream API for efficient file I/O and EF Core with SQLite for lightweight transfer metadata logging

P2P File Transfer - Server-less WebRTC Applic

Ever wished for a personal 'Jarvis' for your desktop? That was the inspiration behind RAVE, a project to build an intelligent and secure voice assistant from the ground up. At its core is a sophisticated natural language understanding module we engineered, combining Whisper-based transcription with an LLM for command parsing. This allows RAVE to go beyond simple, hard-coded commands and respond to user intent in real-time. To ensure it works reliably even in noisy environments, we integrated an RNNoise CNN for real-time noise suppression. Security was a top priority. Unlike many assistants, RAVE features integrated wake word detection and speaker verification to significantly reduce false triggers and enhance user safety. The result is a highly responsive assistant that can manage files, launch applications, and execute system commands, all through natural voice input We are proud that this project was reviewed and accepted into the official winget-pkgs repository. You can install it directly on any modern Windows machine with the following command: winget install Scripter2104.Rave