Classify

AI pilot Full-stack engineer

Document and records classification for institutional workflows — Python ML services with face embedding search and vector-store similarity.

Overview

classify pairs a TypeScript admin UI with a Python backend that handles face detection, embedding extraction, and similarity search. Redis caches hot vectors; Supabase persists institutional records. The split keeps heavy inference off the browser while still exposing real-time WebSocket updates where needed.

What it covers

Face detection pipeline with Pillow preprocessing
Vector-store similarity for record matching
Redis-backed embedding cache
FastAPI routers with structured attendance exceptions

Classification pipeline

import face_recognition
import numpy as np
from utils.vector_store import find_similar_faces
from utils.cache import get_redis_client
from models.face_embedding import FaceDetectionResult

Screenshot placeholder — classification queue with confidence scores.