ARYAN MISHRA
Backend • Systems • Full Stack Web Applications
I'm a 19-year-old, 2nd-year Undergrad Developer who builds software primarily using C++ and TypeScript. My work ranges from networking and systems projects in C++ to full-stack web applications in TypeScript. I am well-versed in the Node.js ecosystem and technologies like Redis, along with infrastructure tools like Docker. Currently, I am deepening my skills in low-level systems programming, distributed systems, and backend engineering.
Featured Projects
A high-performance job scheduler featuring a hybrid queuing architecture (Work-Stealing Deque + MPSC Mailbox) designed for microsecond latency and massive concurrency.
Key Features
- Achieved < 5µs P50 latency and 1M+ jobs/sec throughput via spin-wait spinning optimizations
- Implemented Lock-Free Chase-Lev Deque and MPSC Mailbox for contention-free task injection
- Used a Randomized Work-Stealing algorithm to prevent convoy effects under heavy load
- Zero dynamic allocations on the hot path with cache-aligned structures to prevent false sharing
A multi-process monolith for trimming clips. The Express API offloads heavy yt-dlp + FFmpeg work to BullMQ workers backed by Redis, streaming real-time progress over WebSockets.
Key Features
- Multi-Process Monolith: API Server + Worker
- Real-Time Job Status via Socket.io
- BullMQ Worker System with Priority based queues
- Google OAuth Authentication
- Optimized background job orchestration using BullMQ and Redis, ensuring system stability during burst of 4,000+ concurrent jobs.
A thread-safe, stop-the-world Mark-and-Sweep garbage collector in C++. It manages heap objects via VM stack and global roots tracking with safe multi-threaded allocation.
Key Features
- Tracing-based garbage collection with stop-the-world design
- Multi-root tracking (stack + globals)
- Cyclic reference handling
- Thread-safe heap access using a single global mutex
- Multithreaded stress-tested allocation safety
An AI-powered educational chatbot for Indian history, featuring strict scope enforcement, context-aware continuations, custom stateless auth, and Redis-backed rate limiting.
Key Features
- AI chatbot powered by Llama-3.1-8b
- Context-aware follow-ups using Redis-backed temporary memory using Sliding Window Algorithm
- Atomic Redis Fixed Window Counter Rate Limit Applied
- Rate Limited o allow only 10 chat messages per 60 seconds for LLM API Security and 5 authentication requests per 600 seconds to not allow spam requests across My Server
- Custom stateless authentication (JWT + Cookies)
Latest Blog.
Check out some of the blogs I have written.
Currently Building
Conduit
Conduit is a distributed SQL router built with Node.js and TypeScript. It intercepts PostgreSQL wire protocol packets and parses SQL ASTs to transparently route queries across sharded database instances based on consistent hashing.
Technical Arsenal
Select Category