Aasheesh Rathour
Go engineer building high-throughput systems from first principles. I built a pub-sub broker that beats Redis by 62x, implemented io_uring from raw syscalls in pure Go, and reverse-engineered Bluetooth protocols because I wanted to. Currently building the entire backend at Muze, shipped the MVP in 30 days. Most of my time goes into kernel internals, lock-free data structures, and figuring out why things are slow. I also contribute to open source when I find something broken.
Experience
journey- Architected and delivered entire platform MVP in 30 days using microservices architecture with Kafka, Redis, and Pub/Sub, handling 10,000+ concurrent WebSocket connections.
- Built production-grade real-time chat system with message delivery workflows, read receipts, and push notifications.
- Integrated Mollie and Crypto.com payment gateways with webhook handling and subscription flows processing.
- Engineered advanced search using Elasticsearch, vector embeddings, and Ollama (LLM inference) for semantic search and personalized recommendations.
- Leading full-stack development with Go backend microservices and Next.js frontend, focusing on real-time notifications and scalable infrastructure.
Contributions
View Feed
Selected Works
A high-performance distributed pub-sub broker built in Go from scratch. Outperforms Redis by 62x using lock-free ring buffers and zero-allocation message paths. GitHub•Read the story →
Bypassing fsync bottlenecks by implementing io_uring from scratch using raw syscalls in pure Go. No CGo, no third-party libraries. Achieved 5.5x throughput under concurrent load. GitHub•Read the whitepaper →
An LSM tree storage engine built in Go from scratch. Handled 2M writes/sec, beating BoltDB by 5x on random writes. Features leveled compaction, sparse indexes, bloom filters, and a zero-syscall WAL. GitHub•Read the story →
High-performance Go SDK for the Lamatic AI platform. Built with idiomatic internal/pkg separation, supporting complex AI orchestration through workflow (Flows) execution, Agent interaction, and robust job polling with dual-layer (API Key & JWT) authentication.
LLMs suck at date math. Built this to solve it. A LangGraph agent that handles complex scheduling by separating intent extraction from deterministic resolution. It features robust day-of-week logic, cross-timezone conversion (IST/UTC/Local), participant extraction, and anchor-based ranking to find the perfect slot. Zero hallucinations, 100% accuracy.
Indestructible memory for AI Agents. A high-performance, durable runtime built in Go. Read the story →
AI-powered email template platform. Domain backed by Peerlist founder.
Notch clipboard manager for Linux and Windows. Received 55+ upvotes.
Experiments
A physics simulation in Go where water drops interact with and displace sand particles.
Built a lock-free, multi-producer/multi-consumer ring buffer in Go. Zero memory allocations, 4x+ faster than Go channels, and performance scales with batch size.
Emoji Feedback
Interactive emoji feedback components with smooth animations and haptic feel.
Built a fast, local full-text search engine in Go from scratch. Features 100 concurrent workers for indexing, inverted index building, and fuzzy search (Levenshtein distance) for typo tolerance. Like a tiny, lightweight Elasticsearch.