Creating magic requires
passion and technical excellence Join the team behind it

Looklet technology

Job openings Connect

Our elite team combines excellence in technology with a fundamental passion for photography and fashion. Innovation happens at the intersection of people and ideas--- this is why our office sits right in the middle of the creative pulse of Södermalm, Stockholm. When tech meets fashion, you get 3D printed feet lying around the office, fashion style-offs and daily lunch with colleagues wearing everything from Prada to T-shirts.

Our diverse team of creatives and developers work hand-in-hand with our skilled suppliers of 3D modeling and hardware components. On a joint journey to understand and serve our clients better, we continuously invent new and fascinating solutions. As part of the larger tech ecosystem, we welcome and leverage open source and third party solutions that allow us to focus on what we do best: creating the Looklet Magic.

Looklet and the tech community

Within our proprietary software product, Looklet’s clients can freely mix and match models, body variation and styling items. To create this flexibility, each “base body” is split into 800 distinct digital assets using advanced image recognition and rendering algorithms. This exciting synergy among software development, hardware construction, photography and fashion is attractive to any problem solver or innovator.

CLOUD-BASED BACKEND
Our primary development platform is Java 8, with add-ons for a web front-end using JavaScript and advanced image recognition algorithms in C++.  Our cloud-based backend system is fully automated to scale with client production as well as seasonal variation. As a data-driven company, we make informed decisions based on the in-depth data available to our business analysts via BigQuery and Google Data Studio.

AGILE DEVELOPMENT
We believe in working agile within an agile framework with quick sprints, continuous delivery and a micro-services technology: Jenkins, Docker, Ansible, PostgreSQL, Tomcat, Redis, RabbitMQ, Ubuntu, Kibana, Elastic Search and more.

To support our growth plans for Looklet technology and business in the future, we continue to invest 30% of our current development into our platform, making it “future proof“ for agile innovation.

COMMUNITY MEMBERS
Our strategy is to focus our developers on our unique Looklet technology, therefore, we widely adopt third part services and partner with open source technology to extend our reach beyond our capabilities.  This allows us to firmly grounding our developers in initiative to thrust the development of our product forward, while encouraging our developers to maintain an active role in the technology community.  We believe in sharing our best practices and being influenced by others. Accordingly, we often co-host developer meetups in our office, partner with developer conferences like Jfokus, Oredev, and DevOps Summit, and present our technology at the development conferences, such as our visit talk at JavaOne in San Francisco.

Interesting articles
Using an AWT splash screen with JavaFX on OSX

The Java launcher command line option -splash does not work well together with JavaFX on OSX. A way to work around this is to let your application start a separate Java process that only shows the splash screen.

Expertens bästa tips: så lyckas du med devops

Så länge man skiljer på utveckling och drift så blir det svårt att lyckas med devops. Den och många andra erfarenheter har Cecilia Borg på Looklet gjort under många år med devopssatsningar.

Teleporting clothes from a mannequin onto a model

Interested in how you teleport clothes from a mannequin onto a model? Read Christopher Dalid's post in two parts about this.

Definition of magic
[ˈmædʒɪk]
Possessing distinctive qualities that produce baffling effects.

Software developer at looklet

In the role of software developer at Looklet, you will become part of an organisation of cutting-edge professionals. When working with photo realistic digital production for large fashion retailers, quality solutions are not an aspiration, but a requirement. If the digitally rendered model images don’t look like they were shot live in studios- they are useless. It’s all or nothing. And that is a worthy challenge!

Our strategy to achieve Looklet Magic creates a cross functional development-focused organisation, staffed with highly capable individuals in software development,
hardware construction, photography and fashion. The mix of these diverse backgrounds creates a dynamic environment with an abundance of innovation and, as it happens, a lot of fun. We have agile methods in our hearts and cloud platforms under our feet.

Fashion Tech is an industry in its infancy and Looklet is at the forefront of defining the space as our solution is is quickly replacing the traditional approach across the world. And there is more to come. Join us in building pioneer solutions that change a whole industry!

Research

Looklet’s Research team is on a mission to exploring the vast, multidimensional domain of computer science, machine learning, computer vision, studio photography and digital color representation.

An intimate team of 8, with varied backgrounds and focus, work exclusively within the world of revolutionary and newly developed research that serves as the pillar of our solution. To stay at the forefront of the blossoming fashion-tech industry, we’ve dedicated a whole team to inventing new ways of looking at the intersection of fashion and technology.

Six Research Engineers, one Product Owner and a a Team Manager gather, study and research advanced computer science, image recognition and machine learning topics integral to the Looklet solution.

Two technical challenges

Martin Pettersson

Research engineer

I’m part of the research team at Looklet. The idea with the research team is to find new solutions to existing problems, but also to follow research and new technology that one day can become part of our products.

Right now, together with two other researchers, I’m evaluating algorithms for alpha matting, the code that removes the green backgrounds and mannequins, leaving only the garment. Our current implementation works well in most cases, but sometimes the results could have been cleaner – and even worse, green garments sometimes disappear completely.

There are tons of research papers on the subject, and even off-the-shelf solutions, especially for chroma keying used in TV production. However, since we want to keep green garments, we cannot rely on color alone. The approach I’m currently investigating is based on probabilistic graphical models together with something called Belief Propagation, and so far it seems very promising.

The approach I’m working on, if it holds up, will then be part of the general approach and be combined with other techniques that we have developed and continue to develop. When we have an overall solution we can stand by, a proof-of-concept, we present these findings to the development team and the product owners. Hopefully, my work of filtering, evaluating, and translating, saves the development team valuable time and improves our product at the same time.

Olof Larsson

Software developer

CAMERA INTEGRATION STABILITY
As a software developer at Looklet I code a lot of Java. This involves working on our cloud hosted micro services, but also the control software for our physical hardware studios.

The Looklet studio solution contains a Canon camera. This camera is remote controlled using the open source software GPhoto2, which we launch from within the Java application.

I recently solved a problem where the Canon camera would hang and become unresponsive. As the problem happened intermittently at our clients’ studios, we had to find creative ways to rule out problems with the camera, manual handling, our own code as well as different third party software.

I created a Kibana graph based on our production error logs. This allowed me to see that all studios were affected and not just a few. To pinpoint the cause I rewrote the source code for testability. By breaking out logic to dedicated single purpose classes, applying fail-fast design, and finally coding unit tests, I was able to find and address a couple of stability flaws in the interaction with GPhoto2.

The same Kibana graph from before also served as a follow-up tool. By looking at it we could verify that the problem stopped occurring in production after release.

WHAT'S UP