AI Video Analysis
Scoring long-form video on multi-signal features to find the moments worth clipping.
- Computer Vision
- LLMs
- FFmpeg
Experiments, prototypes and half-finished ideas. This is where I test what's possible before it becomes a product.
Not everything needs to become a startup. Some of this exists purely to find out whether it can work.
Scoring long-form video on multi-signal features to find the moments worth clipping.
Batch pipelines that produce personalised content from a template and a dataset.
Retrieval architectures tested against real document sets, not benchmark demos.
Multi-step agents with tool access, budgets and hard stop conditions.
Small tools that remove friction from my own build and deploy loop.
Glue systems that quietly remove manual steps from real workflows.
Studies in motion, type and generative composition — no product attached.
Repository data is configurable in src/data/lab.ts — no API key required.
ragiq-core
Retrieval pipeline primitives: chunking, embedding, hybrid search.
nest-starter-prod
Opinionated NestJS starter with auth, queues and Docker.
batch-render-engine
Queue-backed batch generation with resumable runs.
prompt-kit
Typed prompt templates with variables and evaluation hooks.
k3s-deploy-notes
Deployment recipes for small production clusters.
ui-primitives
Editorial motion and layout primitives for React.
Contribution activity
Illustrative grid — connect a live feed or edit the data file to make it real.
Have a product idea, an existing application that needs improvement, or an AI workflow that should exist?
Available for selected projects