wastewise
A restaurant uploads its sales history and picks a location and forecast horizon. An XGBoost model forecasts per-item demand, backtested against a seasonal baseline on a 7-day holdout. the improvement is measured on screen, not asserted. An LLM agent then adjusts each forecast for upcoming weather and holidays, giving a reason per item. Next, a sourcing agent prices every item against live retail listings and a US retail average benchmark, picks the best plain-commodity listing, and explains its choice. The result is a drafted purchase order with line totals, savings versus benchmark, and an agent-written purchasing rationale that a human reviews, approves, and exports to CSV. Both LLM judgment steps (demand adjustment and supplier selection) run on an open model served via vLLM on the AMD Developer Cloud, through an OpenAI-compatible endpoint swappable with a single environment variable.
Category tags:
.png&w=96&q=75){kind=small}
.png&w=96&q=75){kind=small}
.png&w=96&q=75){kind=small}
For Innovators & Creators