sahil singh

Backend & AI Engineer | Super 30 3.0 @100xDevs | CF 1841

Software engineer focused on backend systems, AI applications, and low-latency infrastructure looking for AI agent and model-inference roles. At WyiseAI I built an avatar-based LLM chatbot: multi-persona conversations with context compaction, summarization, and memory management for long chats, plus the backend services and AI integrations that keep it running in production. Outside work I go deep on systems: • GPT-2 Small inference engine in C++/CUDA — BPE tokenization through transformer blocks, custom GPU kernels, KV caching; ~0.3 tok/s on CPU → ~20 tok/s on a GTX 1650 • Event-sourced perpetual futures exchange — in-memory matching engine, Redis/PostgreSQL, WebSockets • High-performance slab allocator — up to 5.44× faster than the system allocator Competitive programming: Codeforces max 1841 (High Expert). Meta Hacker Cup 2025 — AIR 38, world rank 377 / 13,779. Stack I use often: C++, CUDA, TypeScript, Python, Node.js, PostgreSQL, Redis, WebSockets, LLM apps / agents / RAG / context management.