Full-stack engineer building real-time, asynchronous systems with applied AI/LLM integration.

Semantask — AI-native work coordination: teams talk, AI extracts the work as suggestions, managers approve and stay in control. Autonomy is optional and policy-gated, not the default.

• 💬 Suggest-first extraction from realtime chat — review before any side effects
• ✅ Manager approvals, tool grants, org policy, and audit trails
• 🔄 Transactional outbox + idempotent, lease-based async task execution
• 🤖 Optional multi-provider LLM execution layer, gated behind explicit policy modes

Product direction: ADR-005

Knowledge Workbench — NotebookLM-style research assistant with hybrid retrieval (dense + lexical, RRF-fused, LLM-reranked) and citation-grounded answers that jump back into the original source.

• Real-time systems (Socket.IO, presence, horizontal scaling)
• Asynchronous processing (outbox pattern, idempotency, distributed leases, retry orchestration)
• Applied AI/LLM integration (agentic execution, RAG, hybrid retrieval)
• Backend & reliability engineering

• LLM-based ingress classification (shadow-mode testing against the current regex classifier)
• Retrieval quality evaluation for RAG pipelines
• Fault-tolerant, worker-pool coordination patterns

• Full-Stack / Backend Engineering roles
• Junior / SDE-I positions
• Startup environments
• Open source collaboration

Suggest first. Approve when it matters. Automate what's earned trust.