SentinelBrain — Mission Control

loading…

Recommendations

—

#	Category	Rows	Tokens	Skipped

Sample rows

Watched

select a log…

mode: —

When no checkpoint exists, this endpoint returns a tokenize-only echo so you can verify the pipeline. Once a checkpoint lands, lazy-loads automatically.

AMD Developer Hackathon 2026

Build AI agents & high-performance AI apps on AMD GPUs

$20,000 prize pool AMD Radeon AI PRO R9700 GPU May 4–10, 2026 lablab.ai

View hackathon page →

Track 2 (our fit)

Fine-Tuning on AMD GPUs

Domain-specific LLMs fine-tuned for accuracy on ROCm. Tech: PyTorch + HF Optimum-AMD + vLLM.

Track 1 (combine)

AI Agents & Agentic Workflows

Intelligent systems that automate workflows, coordinate agents. LangChain, CrewAI, AutoGen.

Extra Challenge

Ship It + Build in Public

Document journey, share insights, provide ROCm feedback. Dedicated prize pool.

Our AMD Infrastructure

Elastic GPU mesh — AMD Instinct nodes

—

Node	Host	Arch	VRAM total	Free	Healthy	Shards

Register a new AMD node (curl)

curl -X POST http://<mesh-master>:7860/api/register \
  -H "X-Agent-Token: $MI300X_AGENT_TOKEN" -H "Content-Type: application/json" \
  -d '{"name":"aios-node-2","host":"10.0.0.12","port":7860,"vram_total_gb":192,"arch":"MI300X"}'

Judging Criteria

Application of Technology

How effectively models are integrated into the solution

Presentation

Clarity and effectiveness of project presentation

Business Value

Impact and practical value in real business areas

Originality

Uniqueness, creativity, novel approaches

Submission Requirements

☑ Public GitHub Repository (MIT license)

☑ Demo Application URL

☑ Video Presentation + Slide Deck

☑ Hugging Face Space under AMD Hackathon HF Org

☑ 2+ social media posts tagging @AIatAMD + @lablabai (Ship It challenge)

Φ Geometric

—

Φ EMA

—

Φ Normalized

—

Trend / Samples

—

Φ trajectory (geometric mean of partition MI)

Φ rising = integrated information growing. Sudden drops trigger self-evolution rejection (gate from §10/§12).