Marvik

BACKGROUND

Please describe your company and position.

I am the CEO of Inno-Aut

Describe what your company does in a single sentence.

We make technology products for the autistic and neurodivergent community.

OPPORTUNITY / CHALLENGE

What specific goals or objectives did you hire Marvik to accomplish?

• Help us implement AI solutions

SOLUTION

How did you find Marvik?

NVIDIA GTC

Why did you select Marvik over others?

• Pricing fit our budget
• Good value for cost
• Company values aligned

How many teammates from Marvik were assigned to this project?

2-5 Employees

Describe the scope of work in detail. Please include a summary of key deliverables.

Project: Chat‑Agent & Activity‑Agent Integration for Aida (2024)
Client: Inno‑Aut, makers of Aida – a therapeutic communication assistant for autistic individuals
Scope:

Architecture & Development: Designed and built a serverless chat‑agent backend in Python (AWS Lambda) that seamlessly integrates with our existing chat agent
LLM Integration: Connected multiple OpenAI models for context‑aware dialogue, dynamic prompt engineering, and fallback handling
Security & QA: Executed a comprehensive security review to ensure user data safety and regulatory compliance
Outcome: Delivered a robust, production‑ready chat service module

Challenges:

• Coordination with an external mobile‑app team whose underperformance and misrepresentation led to delayed timelines.
• Rapid iteration under tight budget constraints, requiring prioritization of core features and security

Technologies: Python, AWS Lambda, OpenAI GPT‑3.5/4, IAM, DynamoDB, REST APIs

RESULTS & FEEDBACK

What were the measurable outcomes from the project that demonstrate progress or success?

Measurable Outcomes & Successes

40 % boost in response accuracy from chat agent itself

Live demo “saved” at Woodland Community College (stakeholder): Though our mobile app wasn’t ready, the Python‑powered chat agent module allowed us to demo the core experience to 10+ college representatives—preserving the pilot opportunity and preventing a canceled meeting.

Stakeholder buy‑in: Following the demo, Woodland CC verbally agreed to a paid pilot—our first institutional commitment, directly attributable to the chat agent’s performance.

These outcomes demonstrate that—even under tight timelines and with external hiccups—Marvik delivered a robust, secure, and high‑impact chat capability that kept our momentum alive.

Describe their project management. Did they deliver items on time? How did they respond to your needs?

Marvik delivered all objectives on time and even provided documentation along with case studys to support the mission itself. We could not have been happier.

What was your primary form of communication with Marvik?

• Virtual Meeting
• Email or Messaging App

What did you find most impressive or unique about this company?

I think what is most impressive about Marvik is there commitment to excellence and love of technology, specifically AI technologies. 

They were always friendly and well-organized as well.

Are there any areas for improvement or something Marvik could have done differently?

I could not ask for a better experience.

BACKGROUND

Please describe your company and your position there.

My name is Marcelo Blanco, I’m CEO & Cofounder at GU Trade.

GU Trade provides mass consumption and distribution companies with visibility of execution at Point of Sale, allowing for evidence-based decision-making and monitoring of the main KPIs. It is a spinoff of Xseed, a Uruguayan software factory.

OPPORTUNITY / CHALLENGE

For what projects/services did your company hire Marvik, and what were your goals?

We approached Marvik looking to automate part of the physical counting process of products across different brands. The main goal was to develop a system capable of determining the share of a given brand’s shelf at supermarkets, using computer vision techniques, based on a picture or a set of pictures. This with the objective of streamlining a task previously done manually.

SOLUTION

How did you select this vendor and what were the deciding factors?

We selected Marvik because of its deep knowledge and experience in the field of machine learning, specifically in computer vision and working with retail clients. More than that, we were very impressed by the fact that they made our problem their own, taking ownership from the beginning in order to solve the problem in the best possible way. Also, we would like to highlight Marvik’s commitment and readiness to help. 

Describe the scope of work in detail, including the project steps, key deliverables, and technologies used.*

While the initial scope of the project was 3 months, it was naturally extended as we identified new outcomes to achieve. 

During the first phase of the project, the selected used case was a coffee brand. Marvik’s tech team started by performing an exploratory data analysis, with the focus on understanding the distribution of the data, generating an intuition of it and determining the type of predictive model best suited to perform these pieces of data. Feature Engineering techniques were used to prepare the data for a data processing pipeline.

Then, they proceeded with the assessment of ​​different machine learning and deep learning models, benchmarking the best possible architectures identified to address the problem.

Once the best model was identified, it was fine-tuned to optimize the metric at issue. Lastly, they put the model into production, where it was tested against new data that in turn could be used for subsequent re-training and continuous learning.

Among the technologies used were Python, OpenCV, Tensorflow, Keras and Pytorch, the system was deployed to Azure.

Deliverables included:

• Machine learning models

• Docker containers

• Source code

• Deployment of solutions

• Documentation

How many people from the vendor's team worked with you, and what were their positions?

The project involved the CTO and a Delivery Manager, plus the dev team that changed over based on the required skills at each moment (computer vision, data engineering, etc).

RESULTS & FEEDBACK

Can you share any measurable outcomes of the project or general feedback about the deliverables?

The main objective, which was the development of the solution to measure the SOS, was achieved as planned. Two of the most significant measurable outcomes were the number of shelves automatically processed por each use case (hundreds), and the number of objects identified with a high confidence level (tens of thousands).

Additionally, during the course of the project, we identified new objectives to keep working together, such as product categorization and photo cropping.  

Describe their project management style, including communication tools and timeliness.

Marvik works with agile methodologies. Work was divided into two-week sprints. Before the start of each sprint, we met with the delivery manager and tech lead to assess the results obtained in the previous one and set new ones for the following. 

Also, we held weekly meetings to revise the objectives set for each week. This allowed us to be in the loop regarding the results and request changes in priorities or improvements to be included in the next development sprint.

Communication was always fluid and transparent, via Google Meet and Slack. We are very satisfied with the way they handled communications.

What did you find most impressive or unique about this company?

Besides being technically good, what we found most impressive was their commitment to achieve the best results possible. We were impressed by their feeling of shared responsibility and the fact that they made our problem their own.

Are there any areas for improvement or something they could have done differently?

We don’t see many areas that need improvement. Something that could be mentioned is that we would have liked to have more direct contact with each of the developers throughout the process.

BACKGROUND

Introduce your business and what you do there.

I’m the director of a SaaS company based in the UK. We offer a subscription service where we create aging and lifestyle effects to show how a person could look depending if they had different lifestyle habits, for example, smoking, drinking, dieting, exercising, stressing, sleeping poorly, and more. As a B2B company, we license other companies in the financial, corporate, health insurance, and beauty market.

OPPORTUNITY / CHALLENGE

What challenge were you trying to address with Marvik?

We needed Marvik to help us find a solution for some of our outdated aging effects. We also wanted them to implement neural and generative adversarial networks in our service.

SOLUTION

What was the scope of their involvement?

Marvik has helped us implement a series of networking services for our aging effects. At first, the major step was finding the right solution that could ballpark for us, in terms of budget, quality and speed of outputting images. We needed to find a solution that was a good balance between quality, cost, and speed. It has since gone from there.

What is the team composition?

We mainly work with Rodrigo (Co-founder & CTO). Paola (Co-founder & CEO) has also helped us on the source of administration and communication side.

How did you come to work with Marvik?

We were mainly looking for a developer who could help us when I found Rodrigo. He explained to us he had a company with a team behind it called Marvik. We've researched and sought with a few other companies, but after a few talks and meetings, we decided to go ahead and hire them. They had done similar work before and that gave us the confidence they could deliver what we were looking for.

How much have you invested with them?

So far, we’ve spent about £8,000 (approximately $11,000 USD).

What is the status of this engagement?

The project has been sporadic, but we’ve started working together in January 2020, and it is ongoing.

RESULTS & FEEDBACK

What evidence can you share that demonstrates the impact of the engagement?

We hadn’t done this kind of project before, but we had a benchmark of quality in terms of competitive application software. We knew that kind of quality was unachievable at the budget that we had, but we’ve been very pleased with what they’ve delivered so far. They give us the feeling that it is only going to get better. Marvik has met our goals so far, and they’ve always delivered on time, understanding our needs, and helping us to compromise along the way. 

Marvik has worked on machine learning and created an MVP for us, so we’ve got something up and running that’s working, which can still be improved. However, this has nothing to do with Marvik’s quality of work. The quality of work is really good. They give good results that we need to build on from now. 

How did Marvik perform from a project management standpoint?

Marvik’s communication and overall performance have been brilliant. They’ve never waited for us to ask the questions. Instead, they’ve always shared with us the project phase, the progress, and any answers needed. It has been really good project management.

What did you find most impressive about them?

Marvik is always looking for a way to solve your problems. Some companies will only deliver what you asked for, whereas they go above and beyond. They always look for a way to improve things, which is an unusual trait. Marvik is willing, enthusiastic, motivated, and communicative, which are all we can ask for. Moreover, their response times have been brilliant, and they’ve gotten back to us very quickly, which is why we chose them in the first place.

Are there any areas they could improve?

I can't think of a way they can improve. They’ve given us what we needed and more. They have been cost-effective and delivered on time.

Any advice for potential customers?

Be as clear as possible with your project. Have a vision for what you're looking for at the end, so that they can understand what it is you're trying to achieve. I think the most important thing to have is clarity on the end goal.

BACKGROUND

Please describe your company and your position there.

I am the Founder & CEO of Resilience International Lab, which is an organization that has the purpose to accelerate resilience in ecosystems.

It focuses on addressing issues of climate change around strategic planning, technical planning and predictive analysis of events, and the application of Building Information Modeling (BIM) technologies for the analysis of infrastructure impacts and adaptation.

OPPORTUNITY / CHALLENGE

For what projects/services did your company hire Marvik, and what were your goals?

The primary aim of this project is to generate insights about how to deal with time series corresponding to different geographic points, using the data collected from monitoring stations for predictive analysis and interpolation.

The project will include an Exploratory Data Analysis to understand the data availability and quality, work will continue training machine learning models, developing a proof of concept to predict water quality components.

SOLUTION

How did you select Marvik and what were the deciding factors?

I attended a webinar given by one of their co-founders on topics of machine learning and right away got interested in reaching out.

After our first conversation, they immediately showed their professionalism, which was one of the deciding factors along with their reasonable fees.

Describe the scope of work in detail, including the project steps, key deliverables, and technologies used.

• STEP 1: UNDERSTAND AND OBTAIN CUSTOMER INFORMATION Task: Analysis of customer data. A first study of the data provided by the customer, analyzing its completeness and quantity. Deliverable: Documentation with the experiments and analysis performed.
• STAGE 2: EVALUATION OF ALTERNATIVES Task: Depending on the data analysis performed, define with the client possible algorithms and solutions to the proposed problem. Deliverable: Document collecting proposed alternatives and preliminary.
• STAGE 3: PROOFS OF CONCEPT Task: Proof of concept. Deliverable: Source code for developments and tests performed.
• STAGE 4: DESIGN THE ARCHITECTURE FOR SCALABILITY OR FUTURE RECOMMENDATIONS Task: Design the architecture to continue deployment. Deliverable: Document proposing architecture design, including technologies use, modules, interfaces, etc.

How many people from the vendor's team worked with you, and what were their positions?

1. Paula Martinez Co-Founder @ Marvik
2. Rodrigo Beceiro CTO & Cofounder @ Marvik

RESULTS & FEEDBACK

Can you share any measurable outcomes of the project or general feedback about the deliverables?

During the present work three different sources of information were used to develop water quality predictive algorithms. The results are considered promising, we obtained predictive results from the three information sources.

Given the datasets available, Marvik developed water quality forecasting algorithms for different tests. The method ARIMA presented the best results when the amount of data was relatively small -for example for the estuary data.

Nevertheless, when there is more information available -for example for ocean data- deep learning methods like LSTM should not be dismissed. For the data available, the generalization of these water quality forecasting methods, Marvik validated through several stations with good results on two different models ARIMA and LSTM.

For the spatial inference, different Kriging methods were tested. Given the data available the best performance was obtained using Ordinary Kriging with a spherical model.

Describe their project management style, including communication tools and timelines.

Marvik follows a very lean project management style. We established a weekly 30-minute call with Marvik to discuss updates and recommendations.

We also created a WhatsApp group to discuss comments/questions/concerns during each week and their response time was quick and precise. They were on time with the timelines initially set and constantly informed us on the progress.

What did you find most impressive or unique about this company?

The professionalism of their co-founders is reflected in the quality of their work. Their vast experience was noticeable from day one as they quickly understood the core concept of the project and went above and beyond by reading and learning about the scientific base of the project to understand the best methodology to follow and interpret the results adequately.

Are there any areas for improvement or something they could have done differently?

I sincerely do not have anything to add.

BACKGROUND

Please describe your company and your position there.

I am the CTO of the company, Paigo, a financial technology firm that specializes in granting consumer credits in a 100% digital way. We currently operate in B2B and C2B modalities in Uruguay and Argentina, and own 3 different brands to tackle different market segments.

OPPORTUNITY / CHALLENGE

For what projects/services did your company hire Marvik, and what were your goals?

A core aspect of our business is to "know our customer", something that is extremely regulated on all markets where we play and we will. Marvik helped us to develop a way to onboard our customers while we collect critical information for KYC and AML purposes in an easy and decoupled way in order to be used in different Customer Journeys (Whatsapp Bot, proprietary Web and Native applications, etc).

SOLUTION

How did you select this vendor and what were the deciding factors?

We found them through referral. They were fit from the very beginning, since they had exactly the expertise and energy we needed to develop the solution. They clearly had what it takes in order to face all the challenges we encountered along the way while developing, deploying and learned throughout the project.

Describe the scope of work in detail, including the project steps, key deliverables, and technologies used.

The core and API system were developed using Golang, and physical persistence using AWS RDS MySQL server instances. For assets processing and validation, we used AWS Rekognition for CV/image processing and for face recognition and matching. For Object Character Recognition (OCR) we used Google OCR API for text mining/detection over images.

There is a second layer of validation components developed in Python and exposed also as micro-services, that tackles the very same tasks of OCR and Face recognition as an alternate mechanism when their corresponding top-level layers return a non-comfortable threshold for a particular validation. AWS S3 Buckets is where we placed all our assets and metadata when performing validations. ReactJS for the Admin console.

How many people from the vendor's team worked with you, and what were their positions?

We have worked closely with the 2 founders, highly skilled and savvy on Machine Learning and Computer Vision techniques and libraries, and we also had a dedicated resource within Marvik for the ongoing support once we were live.

RESULTS & FEEDBACK

Can you share any measurable outcomes of the project or general feedback about the deliverables?

We are extremely comfortable with what the outcome of the project was, the level of support during the project, and the ability to adapt to changes during the project but also when transitioning the production support phase.

With this very same solution Marvik developed for us, we're currently attending 3 different Brands for different segments, with different on-boarding mechanisms (Whatsapp Bot, proprietary Web and Native App) with no problems at all. Scalability was a must requirement, and it worked well in reality.

Describe their project management style, including communication tools and timeliness.

The run the project following an Agile methodology, same as us, which helped to the collaboration between teams. Asana was the tool of choice for project management, and Slack the main collaboration and communication channel since we are also very comfortable with it.

What did you find most impressive or unique about this company?

The level of expertise on cutting edge technologies, but at the same time, the ability to tailor on which use case scenarios apply them, pros and cons versus different approaches, techs or stacks to be used, bottom line, have a good rationale mapped to a business need on every decision made during the process. We were impressed by the excellent communication channels, tone, and the professionalism of the communication itself.

Are there any areas for improvement or something they could have done differently?

Looking back through the path we walked, I don't think I'd do any different, mostly looking at the excellent results as I mentioned earlier.

BACKGROUND

Introduce your business and what you do there.

I’m the managing director of Pi Data Strategy & Consulting. We focus on analytics and solutions.

OPPORTUNITY / CHALLENGE

What challenge were you trying to address with Marvik?

We were looking for a team to develop a web application.

SOLUTION

What was the scope of their involvement?

Marvik developed our web application that consisted of algorithms. The platform monitors drivers, detecting if they’re eating, driving with one hand, etc. Their team built the backend on Python as well as developed a life cycle for the application.

What is the team composition?

There are three people working with us comprising a backend developer, frontend developer, and DevOps person.

How did you come to work with Marvik?

We sponsored an event where they spoke at. Our team reached out to them about the project, and we decided to work together.

How much have you invested with them?

We’ve invested around $30,000.

What is the status of this engagement?

We began working with them around November 2019, and the project is ongoing.

RESULTS & FEEDBACK

What evidence can you share that demonstrates the impact of the engagement?

The app has gone into production, but we haven’t gotten any customer feedback yet. Internally, we are pleased. Every two weeks, they make sprint review documents so we can easily see their progress.

How did Marvik perform from a project management standpoint?

I’m the project manager. They worked with me to plan sprints and conduct follow-up sessions. We use Jira, and our communication is good.

What did you find most impressive about them?

They’re very knowledgeable and experienced. Their team is easy to talk with, and we’ve built a good working relationship.

Are there any areas they could improve?

It isn’t their responsibility, but it’d be nice to have their help on long-term planning.

Do you have any advice for potential customers?

Present your vision and let them give their ideas to help you find the best solutions.

BACKGROUND

Introduce your business and what you do there.

I’m the director of operations for Celistics in Argentina and Uruguay. We are a logistics company for the consumer electronics industry.

OPPORTUNITY / CHALLENGE

What challenge were you trying to address with Marvik?

We started doing a digital transformation effort in 2019, and that comprised several actions: software interfaces with our transportation company, and then we did several enhancements to our WMS system for what happens in the warehouse.

Our customer interaction has increased a lot in the past months and we wanted to streamline those contacts. We were doing everything by email but we wanted to implement artificial intelligence, and that’s how we connected with Marvik.

SOLUTION

What was the scope of their involvement?

The first project we gave them was an email sorter. We received different requests from our corporate client. We needed to get those emails, log them onto spreadsheets, categorize them, and assign them to an agent to work them.

They created a system to read the natural language of those emails and sort them through directly into the spreadsheet. Every time we get a new email with a new request, the request will be categorized and logged.

What is the team composition?

We had contact with two: Paula (Co-Founder, Marvik) and Rodrigo (Co-Founder, Rodrigo). They came to the meetings and took into account what we asked them for. They were there when we needed to adjust the algorithm as well. We went back and forth with them. They also have their development team.

How did you come to work with Marvik?

In addition to my role with Celistics, I teach at the local university, in the MBA program. I was a tutor and a mentor for business plans, and I mentored Rodrigo and Marisol.

In October of last year, I ran into Rodrigo at an event and heard about Marvik. They came to Marvik, I told them about what we were doing, and that’s it.

How much have you invested with them?

We spent $7,400 with them.

What is the status of this engagement?

We worked together on this project from November 2019 until January 2020. We also have another project going with them to build a chatbot.

RESULTS & FEEDBACK

What evidence can you share that demonstrates the impact of the engagement?

The algorithm freed up so much time for us, now that we don’t have to manually sort emails. When we rolled it out, and we had to give them feedback about the algorithm, they were very responsive. They provided very good support once the program went live.

How did Marvik perform from a project management standpoint?

They have a way of working that went really for us. They work in sprints, where they’d work for two weeks and then give the MVP to us for testing and feedback, and then we’d give it back to them.

What did you find most impressive about them?

They handle the basics. You can tell that they know what they’re talking about. They were technically knowledgeable and could communicate with our client’s technical team. Their presentations and proposals were very professional. They looked really good.

Are there any areas they could improve?

They really need to share what they’re doing with more companies and grow. They have something going on here and need to get customers outside of Uruguay. We have a very small market here. I see a lot of opportunities for them.

BACKGROUND

Introduce your business and what you do there.

I am the Manager of the AI Lab at Microsoft, a platform in big tech within the information technology industry.

OPPORTUNITY / CHALLENGE

What specific goals or objectives did you hire Marvik to accomplish for your AI Lab at Microsoft?

We hired Marvik to provide machine learning engineering services for rapid prototyping.

SOLUTION

What was the scope of work Marvik handled for your project at Microsoft’s AI Lab?

Marvik delivered 200 prototypes for different companies that came into the lab for AI engineering services, working across a variety of areas.

How did you find Marvik?

Referral.

Why did you choose Marvik over other options you considered for your AI Lab at Microsoft?

We selected Marvik because of their high ratings, proximity to our geographic location, pricing that fit our budget, great culture fit, good value for cost, and alignment with our company values.

What is the status of this engagement?

The project started in December 2023 and is ongoing.

How many teammates from Marvik were assigned to your AI Lab project at Microsoft?

6-10 people from Marvik were assigned to the project.

RESULTS & FEEDBACK

What were the measurable outcomes from your project with Marvik that showed progress or success for your AI Lab?

We completed and accelerated the delivery of 200 prototypes, which demonstrated significant progress and success for our AI Lab.

What was your primary form of communication with Marvik during the project?

We communicated through in-person meetings, virtual meetings, and email or messaging apps.

How would you describe Marvik’s project management?

Marvik delivered on time and maintained great communication throughout the project.

What did you find most impressive or unique about working with Marvik on your AI Lab project?

I was most impressed by their great communication, excellent engineers, and their flexibility and responsibility.

Is there anything Marvik could have done better or differently during the project?

No, we grew the project together.