What does Formulation Labs actually do?

Formulation Labs designs, builds, and governs better systems for acquisition, follow-up, reporting, and operations.

Where do engagements usually start?

Most engagements start with a Systems Audit so the real leverage and sequence of work become obvious before implementation begins.

What is the difference between Systems Build and Systems Oversight?

Systems Build installs the approved operating layer. Systems Oversight governs and improves that layer over time so it does not drift.

AI & automation

AI pipeline that turns saved videos into a searchable library

We built a tool that watches the short videos you save on your phone and turns them into a searchable, summarized library you can actually use later.

A tool that captures the short videos you save, transcribes and analyzes each one with several AI models, and turns them into a searchable, summarized library with a weekly digest.

Context

This one we built for ourselves, to prove out a pattern we bring to client work. You save a lot of short videos, and they vanish into a feed you never revisit. We built a pipeline that captures each saved video with one tap, transcribes the audio, reads the visuals, scores it, and files it into a searchable library, then emails a weekly summary of what you saved. It is the same kind of AI pipeline we build for clients, run end to end on our own.

Constraints

Capturing a video had to be one tap from the phone, with no app to open and no friction.

Each video had to be transcribed, watched, and understood, which means coordinating several AI models, each good at a different part of the job.

Search had to understand meaning, not just keywords, so a video can be found by what it was actually about.

It had to cost effectively nothing to run at personal volume.

Work performed

Built a one-tap capture from the phone's share menu that hands each saved video to a background pipeline, with no app to open.

Coordinated several AI models end to end: speech-to-text for the audio, a vision model to read the frames, and a reasoning model to summarize, tag, and score each video, with automatic fallback if one provider goes down.

Made the whole library searchable by meaning using vector search, so a video can be found by what it was about rather than its caption.

Grouped the saved videos into topics and sent a weekly digest of the highlights, all on free-tier infrastructure that costs effectively nothing at personal volume.

Outcomes

1 tapCapture to library

One tap from the phone's share menu sends a video into the pipeline, with no app to open and nothing to file by hand.

Multi-modelAI, end to end

Speech-to-text, a vision model, and a reasoning model work together, with automatic fallback, to transcribe, read, summarize, and score each video.

SearchBy meaning, not keywords

Vector search makes the whole library findable by what each video was actually about, not just the words in its caption.

← Back to work Apply