Skip to content
Qamba Labs

Qamba Labs

We build AI products end to end and run them in production.

An AI product studio.

Bira Events on an iPhone behind it, showing a festival lineup with set times.
SongScroll on an iPhone, tilted toward the viewer: a discovery dial set to Familiar above a grid of recommended songs.

01Music discovery app for iPhone

Live on the App Store

SongScroll

Music discovery that learns from every save and skip.

SongScroll is a subscription iOS app with two ways in: an infinite grid of AI recommendations and a vertical feed of 30-second previews.

A discovery dial sets how far from your taste it reaches, from Familiar to Insane. The recommender learns online, so every save and skip changes what comes next.

5.0 on the App Store

  • SongScroll’s Vibes grid: the discovery dial set to Familiar above large album-art recommendations.
    Vibes
  • SongScroll’s Feed: a 30-second preview of “Letters from July” with buttons to save it or play the full song on Apple Music or YouTube.
    Feed
  • SongScroll’s Collections: saved songs grouped into Liked Songs, Late Night Drive, Morning Coffee and Weekend Energy.
    Collections

02Event platform for organizers

Live

Bira Events

Every event gets its own app. Attendees open it from a link.

Bira is a multi-tenant event platform. Each event runs on its own subdomain with an installable web guide: the map, the lineup, announcements and a moderated photo feed.

Organizers upload the schedule PDF they already have. An AI import turns it into a structured schedule, which they check before it goes live.

The Bira Events website on a MacBook: “Your event. In everyone’s pocket.” next to a phone showing an event map.
  • Bira’s attendee guide on an iPhone: a festival map with pins for two stages, food, first aid, restrooms, water refill and parking.
    Map
  • Bira’s lineup: Saturday on the Sunset Stage, with set times and a heart to save each act.
    Lineup
  • Bira’s community photo feed: an attendee’s festival photo with likes and comments.
    Photo feed

Screens from the public attendee demo.

  • Per-event subdomains
  • Installable web guide
  • AI schedule import, PDF to structured schedule
  • Moderated photo feed

First client

African Street Festival

September 2026

A festival of an estimated 50,000 attendees.

03AI video production for animators and AMV creators

In beta

Qamba Studio

Same character, every shot.

Keeping a character consistent from one shot to the next is the hard problem in AI video. Qamba Studio is built around it.

Staged LLM director agents plan each shot. ComfyUI and MiniMax render the takes. A vision model reviews every take before it reaches the edit.

  1. 01 Plan

    LLM director agents

    • Writer
    • Cinematographer
    • Choreographer
  2. 02 Render

    ComfyUI and MiniMax

  3. 03 Review

    A vision model checks every take

Qamba Studio, recorded in the app

How we build

Specs first. Agents build. Every step gets reviewed.

One command, spec to release

Each product starts as a written spec. Autonomous coding agents take tickets from there to a release, through automated test gates and a human approval. One command runs the whole path.

A reviewer on every generative step

Tests review the code. A vision-model judge reviews the video. Live engagement data reviews the marketing, through HELM, our multi-brand marketing system.

How a change moves from spec to release, and back:

  1. Spec: Written first
  2. Agents build: Ticket to code
  3. Gates: Tests and evals
  4. Human approval: A person signs off
  5. Release: Ships to users
  6. Metrics: Feed the next spec
  7. Metrics from each release feed back into the next spec.

The studio

An autonomous venture studio.

Give it a list of ideas. It researches each one, scopes it, builds the MVP, makes the marketing assets, connects ads, socials and analytics, and ships to the app stores.

Proof so far

SongScroll and Bira both shipped with the build fully autonomous. Research and go-to-market are still run by hand.

Next: closing that loop.

The pipeline today · 2 of 6 stages run without a person

  • Autonomous
  • In progress
  1. Agents build the MVP from the spec, behind test gates and a human approval. SongScroll and Bira were built this way.

The lab suite

The tools that run it.

Four internal apps, one for each job the studio repeats. Every one has an assistant that reads first and asks before it changes anything.

Screens show sample data.

  • Build

    Dev Studio

    Takes a request from ask to release. Agents plan the work, fix it and review it, and a person approves before anything merges or ships.

    Dev Studio run view for a Bira Events ticket: the full timeline from request, plan, fix and two review rounds to approval, beside a phone-width preview of the staged build.
  • Market

    HELM

    Marketing for every brand from one place: brand DNA, campaigns, product photo shoots, a posting calendar and measured results. Drafts post only after approval.

    HELM Studio photo shoot: the SongScroll app on an iPhone in four generated scenes, each with its cost and a keep-for-generation option.
  • Legal

    Accord

    Legal work for each venture: answers that cite the venture’s own saved documents, drafts from licensed templates, and reviews kept beside the originals.

    Accord answering whether SongScroll’s privacy policy is ready for the App Store and California users, with numbered sources and the cited policy open beside the answer.
  • Finance

    Ledgerly

    The studio’s books: every entry with its source, budgets held before work starts, and provider usage set against each venture’s revenue.

    Ledgerly home in a sample workspace: net movement, money in and out, available budget after reservations, a six-month chart and items needing attention.

Humans stay in the loop on anything client-facing or that spends money.