Skip to content

MuskarTech Labs · Research

The science behind our products.

Labs is where questions become experiments. We form a hypothesis, build the smallest thing that can test it, measure it honestly, and only then turn what works into a product.

01 · AI

Visual product search

Applied machine learning that ends up in real software, with a focus on computer vision.

Beta

A computer vision model that finds which product you are looking for from an ordinary photo.

Inside the AI Lab
Illustrative visual. Illustrative visual search: a product detected in an ordinary photo and matched to similar catalogue items.

02 · Security

CipherPass

Small tools that make good security habits easier to keep.

LiveFree · Chrome extension

A Chrome extension that generates a strong, unique password for each site from your master key and the site name. Nothing is stored and nothing is sent anywhere: the same inputs always produce the same password.

Chrome Web Store
Illustrative visual. Illustrative CipherPass flow: a master key and a site name go through SHA-256 and come out as a unique password.

03 · Experimental tech

Drone from scratch

Hardware and other experiments. We build things from scratch to understand how they really work.

In development

We are designing and building a drone from the ground up: the frame, the flight controller firmware, the sensors and the software that flies it. It is our way of learning hardware, control systems and on-device vision end to end.

Illustrative visual. Illustrative top view of the drone in development: frame, four rotors, flight controller and sensors.

Method

How research works here

Research fields

  • Computer vision
  • Retrieval & embeddings
  • Efficient inference
  • Applied cryptography
  • Control systems
  • Embedded hardware
Research notes on the blog
  1. 01

    Question

    Every project starts from a question one of our products, or our customers, actually raises.

  2. 02

    Experiment

    We build the smallest prototype that can answer it, and write down what would count as success.

  3. 03

    Measure

    Results are judged on held-out data and real use, not on a good-looking demo.

  4. 04

    Transfer

    What works moves into a product. What we learn, including what failed, goes into our research notes.