retail-ai

Retail AI benchmark for choosing the economically right LLM by workflow

面向连锁服装门店的B端AI导购助手，覆盖商品与库存查询、高风险审批、Bad Case闭环和区域运营分析。

An agentic AI retail assistant that combines natural language search, visual product discovery, tool calling, conversational memory, and structured product retrieval to deliver intelligent shopping recommendations.

LLM-based retail assistant using Groq API, LangChain, and Streamlit.

Much more than a chatbot: a smart AI retail expert agent that diagnoses problems, verifies part compatibility from a number or photo, and recommends what to buy next.

Computer-vision checkout system using YOLOv8s, FastAPI, React, and TypeScript with validated end-to-end checkout evaluation.

Palette Proof — YouCam-powered personal color intelligence that ranks real apparel colorways and proves the final two with virtual try-on.

Zero-shot SKU onboarding for edge retail AI — single product photo → trained detector in <10 min via LLaVA extraction, BlenderProc2 domain randomization, YOLOv8n + EWC continual learning, and async FastAPI.

A synthetic-data retail AI agent prototype for product discovery, order support, and assistant workflow evaluation.

AI-powered party planning and shopping agent built with Google AI Studio's Vibe Coding and Gemini, featuring real-time budget tracking, a conversational assistant, and hands-free voice control.

Semantic grocery search engine using ChromaDB + SentenceTransformers — find products by meaning not just keywords using cosine similarity HNSW indexing

Real-Time Retail Customer Intelligence — OAK-D stereo camera, MediaPipe, Kalman tracking, CNN embeddings, QAT int8, HDF5, Tkinter dashboard

Retail shelf audit demo comparing Cerebras Gemma 4 inference against local Gemma on Mac.

Computer vision model to estimate customer age from photos. Uses fine-tuned ResNet50, achieves MAE 6.37 (<8 target). Enables personalized offers and age verification for alcohol sales.

Why the biggest AI opportunity in retail isn't at headquarters — it's on the store floor. GRADE Framework: 11 failure patterns, evaluation sequences, and scoring rubrics for Store-Level AI agents.