Terminal-based autonomous coding agent integrating Gemini and Groq models with filesystem and shell tools through a unified streaming interface. Features an autonomous read-act loop with context compaction, session persistence, and provider-agnostic tool execution for interactive and batch workflows.
Hey, I'm Mrityunjay
CSE • BIT Mesra • AI Engineer • 5x Hackathon Winner
Loading…
About
An engineer and open source contributor specializing in AI Systems and Full-Stack Development.
Experience
- Project: Borgbackup Pack Files for borg2 implementing pack file support for the next-generation borg2 storage backend — view all PRs

- BorgBackup (Borg): Authored 33 PRs optimizing CI/CD pipelines, Python 3.12 compatibility, and resolving complex timezone bugs
- BorgBase (Vorta): Authored 16 PRs optimizing CI/CD pipelines, Python 3.12 compatibility, and resolving complex timezone bugs, completing 2 full projects
- OSSF (CVE Binary Tool): Resolved critical Windows crash loops involving sqlite3 databases via filesystem optimizations
- Also contributed to: Supermemory and Hermes Agent (Nous Research)

- Engineered a scalable RESTful API suite and credit-based billing microservice, successfully handling 500+ concurrent transactions using Node.js and FastAPI
- Architected a multi-tenant notification system across WhatsApp, email, and web portals using an event-driven architecture with message queuing
- Optimized PostgreSQL schemas (118+ tables) via advanced query tuning and indexing, improving OCR accuracy by 25%

- Contributing to product development and engineering initiatives as an Associate Software Engineer
- Developed features for the official academic platform to streamline academic management for students, faculty, and administration
- Contributed to building a comprehensive system for managing academic workflows at BIT Mesra

- Built an AI aggregator platform integrating LLMs, MCP tools, local file systems, and image generation APIs using Node.js and Next.js
- Implemented multi-model support enabling seamless switching between Gemini, GPT, Claude, and other LLMs through a unified interface
- Reduced model inference latency by 30% by implementing Redis caching strategies and asynchronous processing queues

- Contributed to software development initiatives enhancing operational efficiency and scalability for an innovative scrap recycling platform
Projects
Distributed workflow orchestration engine that executes DAG-based jobs across Kubernetes runner pods with parallel scheduling and Redis-backed queues. Implements heartbeat monitoring and watchdog-based failure detection, guaranteeing deterministic workflow completion and automatic handling of pod failures.
Centralized cryptocurrency exchange supporting spot and perpetual markets with a high-performance in-memory matching engine, price-time priority, and real-time WebSocket updates. Built on a 6-service event-driven architecture using Redis Streams, WAL + snapshot recovery, and PostgreSQL audit trails for durable order processing and sub-second restarts.
GitHub PR analytics dashboard with 2.5k+ active users visualizing merge rates, contributor stats, and activity timelines for any repository. Features branch filtering, customizable time ranges, CSV/JSON data exports, GSoC 2026 organization lists, and theme support.
Achievements
Skills
Contact
Got a project, idea, or just want to chat? Let's connect.