AtlasLLM

Built a multi-model AI chat platform with a Turborepo monorepo using Next.js 14, Convex, Cloudflare R2 and Redis, supporting 30+ LLMs across 9 providers through a unified OpenRouter abstraction, generating over $3,000 in profit.

Engineered a custom DAG-based workflow orchestration engine powering agentic research pipelines (deep research, pro search) with parallel task fan-out, typed event-driven state, retry/timeout policies, and Langfuse tracing per generation.

Integrated Clerk for authentication and subscription billing, atomic Redis Lua scripts for thread-safe credit deduction, and a production observability stack spanning Sentry, PostHog, Axiom, and OpenTelemetry for error tracking, product analytics, and distributed tracing.