Ahan Phatak@ahanphatak

NodeDash turns a short admin interview into a company's live operating graph. Every node is a department rendered as a box on a hand-rolled, dependency-free Canvas physics graph, paired with an AI agent acting as its Chief of Staff. Edges carry approvals, materials, or data; agents negotiate along them, so a shift ripples through as a renegotiation, not a deadline surprise. The graph is public to view; clicking a dot opens a login modal where an access code claims a scoped seat (or admin opens everything), and the department docks open as a tab: responsibilities, workflows, documents, and a live agent chat. Auth is dependency-free (stdlib PBKDF2 + HS256 JWT). Onboarding is a 4-tier adaptive interview: macro defines the graph, micro defines nodes, workflow defines edges, AI agency defines guardrails. Generation self-heals, invalid LLM JSON gets one repair pass, then falls back to an offline mock so a demo never dies on a flaky model. Example: a vendor slips a checkpoint, its agent recomputes ETA and renegotiates with Procurement — updates ripple visibly, logged for audit. Every agent has RAG memory: similiarly-matched past exchanges, shown in chat as "recalled N past exchanges." One OpenAI-compatible client swaps between Ollama, Fireworks, and a self-hosted AMD MI300X (ROCm+vLLM) via one env var. (An optional production mode pairs an always-on fallback (Fireworks or DigitalOcean Gradient) with the on-demand MI300X it answers instantly while the GPU droplet wakes, then a watchdog destroys it after idling, so cost is zero when unused.) Built, not mocked: 39 offline smoke tests plus a RAG test, dependency-free auth, live graph editing, on-demand GPU orchestration, and a seeded workspace ~4,000 lines. Targets mid-market operators with vendor coordination. Seat-based SaaS per node, one workflow wedge before expanding. Website link - https://fisheries-bond-catalog-matrix.trycloudflare.com

To participate in the hackathon, simply click Sign up with AMD. If you are not already a member of the AMD AI Developer Program (ADP), you will need to create an account. ADP members can access AI tutorials, experts, and community support. In addition, new members can claim a free month of DeepLearning.ai pro, $100 in AMD GPU cloud credits, and $50 in Fireworks AI API credits. All hackathon participants will receive on hackathon launch additional compute and API credits for their projects. Ready to build what's next on AMD? Sign up with AMD below to secure your spot.