By Rodrigo Beceiro

Who are we?

Marvik is a hands-on machine learning consulting firm. We do tailor developments for companies who need to solve a complex problem using AI but can’t do it internally. This usually means they need results faster than it takes to build an internal team or just can’t get enough skilled people.

This article is for anyone who would like to know more about us, but it is specially oriented for those considering applying for a job with us. If you have already made up your mind, just apply here or email us at [email protected]

 

What problems do we solve?

We like to say that we know how to use machine learning to solve problems that no one else has solved before. This is because our clients have very specific needs, related to their unique data and processes. A short list of the problems we have solved includes:

  • Recommender engines at large scales: recommending a specific candidate or item between a set of hundreds of millions of records in milliseconds given a search query
  • Predict why and when will a person go back to the emergency room
  • Predict stock shortages in stores
  • Optimize email campaigns to maximize revenue
  • Predicting house pricing
  • Structuring natural language: Extracting and normalizing attributes from millions of items into a standardized form where it is easy to query, filter, and order
  • Fault detection & prevention: Using computer vision to identify which manufactured parts have failures or raw materials can cause a failure in the production line
  • Automatic summarizing & insight extraction: turning documents of hundreds of pages into a few important points for the reader
  • Exercise detection and repetition counting using computer vision and custom made hardware

 

The team

At Marvik you will find a diverse, fast-growing, and collaborative team. We have extremely talented people that together solve high complexity problems in incredibly fast time. This is the definition of a high-performance team. 

team

People in the team usually value:

  • the learning opportunities
  • always having someone to ask questions
  • the diversity of people and problems they can work with

The mission of every team member at Marvik is to:

  • Look after and motivate people
  • Upper (even more) the technical standards
  • Make Marvik the best place in Latam to learn how to work in machine learning
  • Be professionals when innovating

 

What can you find in terms of learning?

It is of course challenging to constantly be solving new problems but there is a lot of excitement in that as well. Being on the state of the art means that you need to be constantly learning as it is constantly moving and in the machine learning field, at an incredibly fast pace. We are always looking out for new relevant research papers, Github repositories, libraries, tools, and more. This is part of our job to be able to make the most updated and informed recommendations to our clients. 

 

What does our day-to-day look like?

Projects and work methodology

Team members are usually working on one or a maximum of two projects at the same time. 

We have high-level plans which guide us in the longer term and we plan one or two-week sprints. We encourage autonomy and fluent and effective communication (this is how we leverage our team’s experience). 

Dailys are a part of our ritual and help us have a close team, despite working from different countries. They have a strong social component but also help identify blockers and find synergies. We always have someone we can go to if we need help.

The tasks will of course depend on the role and stage of the project. At the beginning of projects, we usually have more research work (Google + Arxiv) and run multiple experiments (we also get our hands dirty with the data).

This initial phase quickly turns into developing an MVP or POC (closer to what traditional software looks like – we replace Arxiv for Stackoverflow). After a first production release, we look to scale whatever we have built. 

 

Tech

We are a technical company, and we can’t write a blog without talking about tech. 

Working with us you would be working with AWS, Azure, and GCP depending on the project and client. Our stack includes of course Python, Tensorflow, Keras, and Pytorch. Additionally, we like to work with Golang and React for any additional full-stack development. Docker is usually a must and the DevOps tools usually change depending on the client.

Stuff we will invest time on in the short term includes:

  • MLOps: it’s just starting. Read more about it 
  • Edge computing: NVIDIA Jetson devices, Google Coral, etc will be more and more common. You could say that machine learning in meeting IoT
  • GAN: Their applications go way beyond facial manipulation with StyleGAN and synthetic dataset creation, they are one of the most creative ideas and always fun to work with
  • Transformers: We think we started early with them and their potential never ceases to amaze us. We have used them for human resources, e-commerce, and health applications with excellent results. On a separate note: special thanks to Hugging Face for what they are open sourcing to the community.
  • The next transformers: we are still not sure what that would be, but we are looking for the next novel architecture that will revolutionize the field 

 

Work-life balance

We encourage all team members to have a healthy work-life balance. We work hard during the business hours, we rest after that 🙂 We could argue that many studies prove that this leads to better-rested minds and increased productivity, but to be honest we do it because people are just happier in that way and that is just better for everyone.

 

Growth

We understand that anyone who joins us will need to learn and grow, and we plan for that. We are constantly looking to improve our onboarding process and have the mission of being one of the best places to be learning and working with machine learning.

Needless to say, we look for long-term relationships in which people can project themselves growing. Marvik is growing at a fast pace and we need people to not only accompany that growth but also push it. 

We are small enough for every new hire to make a difference, and large enough for everyone to be able to do what they do best.

 

Relax

With all this hard work, we also need time to relax. That’s why we combine all of the above with this:

We celebrated Halloween and of course, we dressed up for the occasion.

 

During our end-of-year getaway at Punta del Este, we discovered excellent swimmers.

 

When it comes to celebrating, our team can organize after office in a blink of an eye.

 

We are also great hosts and organized city tours of Montevideo and Punta del Este with part of our Argentinian team members.

 

#hiring

We hope to have given you an overview of what Marvik is about. If you have any additional questions just let us know! Drop us a line at [email protected] and check out all the open positions we are looking for here.

Shape
Get in touch with one of our specialists. Let's discover how can we help you.
Training, developing and delivering machine learning models into production
Document