Sentinel Generalist:

Sentinel Generalist is a zero-shot agricultural intelligence system powered by a large Vision-Language Model (Qwen2.5-VL-7B-Instruct) running on an AMD MI300X GPU. Upload a single photograph of any plant — indoors, outdoors, healthy, or dying — and Sentinel performs 11 simultaneous agricultural analyses in real time, streaming its step-by-step reasoning trace like an expert agronomist thinking out loud. Reasoning trace — watch the AI examine shadows, soil texture, leaf color, and turgor in real time Species identification with confidence scores (50,000+ species, zero-shot) Geospatial inference — USDA hardiness zone, latitude, climate class deduced from shadows and soil Light assessment — estimated daily sun hours + adequacy check Watering analysis — visual turgor + soil moisture cues Nutrient deficiency — N-P-K + micronutrients with organic remedy doses Pest & disease — IPM-first diagnosis (organic → chemical → prevention) Companion planting — recommended neighbors + antagonists + placement instructions Harvest intelligence — days to harvest, visual cues, succession planting Seasonal planning — frost risk, crop rotation, next-season prep Beginner garden planner — "What should I plant right now?" with 7 curated plants for your zone and season Garden harvest preview — interactive top-down layout + AI-generated image of your garden at peak harvest Why AMD MI300X? The 192GB HBM3 VRAM loads the full 7B model with room to spare. ROCm 7.0 + gfx942 optimization delivers ~23% faster inference than comparable cloud APIs. Your plant photos never leave the Droplet — privacy-first by design. The demo differentiator: Tech stack: Qwen2.5-VL-7B-Instruct · AMD MI300X · ROCm 7.0 · HuggingFace Optimum-AMD · FastAPI · React 19 · GitHub Pages · NDJSON streaming · 100/100 Lighthouse · zero trackers · no CDNs