I built a local Elixir/Python pipeline to curate 14,000 RAW photos

Name: I built a local Elixir/Python pipeline to curate 14,000 RAW photos
Availability: InStock
Author: qweliantanner

by qweliantanner·Apr 20, 2026·5 points·2 comments

Visit Project View on HN

AI Analysis

●●SolidNiche GemBig Brain

CLIP + Ridge Regression beats LLaVA for learning personal taste from ratings.

Strengths

•Smart pivot from vision LLM classification to preference learning from behavior
•Elixir/Phoenix orchestrator with Python/FastAPI AI workers is thoughtful architecture
•Handles RAW, TIFF, RAF formats with local inference — no cloud dependency

Weaknesses

•Personal workflow tool — unclear if it's packaged for broader use
•Photo curation already served by Lightroom, Pixea, and similar tools

Post Description

I had 14,000 photos sitting on a drive and wanted an excuse to play with local vision models and Elixir/Phoenix. I originally tried to get LLaVA to tell me if a photo was 'good' or matched my style, but quickly learned that LLMs have terrible taste. I ended up demoting the LLM to just extract metadata, and built a custom CLIP/Ridge Regression pipeline to actually learn my preferences based on how I rate things.

The stack is Phoenix/Oban on the orchestrator side, and Python/FastAPI/Instructor for the AI workers. Happy to answer any questions about the architecture, fighting with local RAW file ingestion, or the pains of Pydantic validation with open weights.