Numlabs

Introduce your business and what you do there.

I’m the CEO of Apriside; we’re a software company that provides for organizations in the oil and geothermal industries.

What challenge were you trying to address with Numlabs?

We were looking for a company that could support us in AI and machine learning.

What was the scope of their involvement?

Numlabs helps us in the development of a few enhanced functionalities for AI and machine learning software. They work together with my team of developers, doing general coding, frontend development, and logic. In terms of technologies, they used different Python libraries to create machine learning algorithms based on the data and requirements we provided to predict sequences of steps for planners.

What is the team composition?

We’ve worked with 3–5 people from Numlabs — the team members vary depending on the stage. We’ve had testers, project managers, and developers, but our main point of contact is their CEO.

How did you come to work with Numlabs?

They were referred to us by one of our developers. We looked at other companies, but we knew Numlabs’ work, and there was some trust in place.

How much have you invested with them?

We’ve spent $60,000–$70,000 with them so far.

What is the status of this engagement?

We started working together in November 2019, and the engagement is ongoing.

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

The most important part for me is that the software works; our customers are impressed by its functionality and how it helps them. There have been parts that are hard to describe when it comes to engagement, and not everything works perfectly fine from the beginning, but Numlabs has done all adjustments and fixes without additional cost. 

How did Numlabs perform from a project management standpoint?

I have no issues with them. They’re always well-organized, inform us on time about new situations, and transparent about why and what — they deliver on time and cost. I’m delighted with that, and it’s why I’m still collaborating with them.

In terms of communication, we were using Slack when the collaboration started, but we’ve switched to Rocket.Chat, along with phone calls.

What did you find most impressive about them?

The communication is accessible, they have a good understanding of our needs, and are very skilled and professional in what they do.

Are there any areas they could improve?

I don't think there's anything they can improve on.

Do you have any advice for future clients of theirs?

I would recommend them; I’ve done it already, and everyone seems happy with the collaboration. The client can always call and ask Numlabs some questions, and they’re good at explaining and helping — that helps create a friendly environment.

Introduce your business and what you do there.

I work at a digital production studio. We're working in the marketing industry primarily and the types of work that we do are different kinds of cross interactive websites, mobile apps, and also experimental technology, and virtual reality. I hold the position of technical director.

What challenge were you trying to address with Numlabs?

I'll keep the brand and the project name hidden because the project has not been released yet — but I'll try to give you as much detail as possible. The project involved a machine learning challenge and it was actually quite an interdisciplinary job that involved analyzing video footage from a famous director.

We used machine learning technology to identify certain characteristics that were specific to the director and their old films, to create a new film that resembled their style.

What was the scope of their involvement?

Numlabs didn't work on the video production, they only worked on machine learning algorithms that were used to analyze the old footage. Some of those algorithms were also used to produce concluded findings and statistical data, that informed the film production process that happened next — they weren't involved in the film production process itself.

They mostly provided custom software development services to provide and analyze data. As a machine learning company, they work with certain market standard technologies like Keras and TensorFlow. Those were the technologies that then served them to develop custom solutions for us, and custom machine learning algorithms to analyze the data that we had. The problem that we had analyzing the archival footage was actually known standard, I would say. 

There was a lot of custom software development that they had to put in order to complete the challenge. I should also mention, that they didn’t only analyze video footage, they also analyzed the scripts of the movies. In terms of the machine learning services they provided, it was both vision and also natural language processing, essentially analyzing written text.

The project actually lasted for several months as we finished the collaboration with Numlabs almost a year ago. We worked together for over six months, I believe, and the project consisted of phases. They started by analyzing EDR archival footage, produced some timings, and then the next phase was more on coming up with what the new movie could look like using those findings. Finally, at the end of the project, they focused on developing technical ways of testing the output just to see how close it actually was to the style of that director. So we can say that the project was run in phases more or less.

What is the team composition?

We worked with around three people. The number could have varied between 2–4 depending on the stage of the project. First of all, we were in touch with one of their project managers. It was mostly on just like the organizational side of things, but it wasn't someone who was actually managing the project on our side — it was more like just a point of contact on their company side. We worked directly with three of their machine learning engineers.

How did you come to work with Numlabs?

I believe it was through a direct search online. We were basically looking for a machine learning vendor and we found their website, which seemed to be essentially describing who we needed — we reached out to them and that's how it started.

How much have you invested with them?

I'll give you an approximate number. It's not going to be super precise, but it was probably around £30,000–£40,000 GBP (approximately $41,400–$55,300 USD).

What is the status of this engagement?

We started working together around March 2020 and the project ended in September 2020.

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

The quality of their work was definitely good. I believe you've got a starring system where you give the companies points on a scale of five so I'll try to use that to maybe quantify my feedback. In terms of the quality of their work, I think the quality is close to five — it could probably be a five in that regard. They were professional and in a nutshell, we were very happy with the collaboration. The only negative, I think, that we ever have raised with them was the ability to deliver on time. 

They were always explaining that the development was a very challenging topic and a very fundamental one, and it was done on a scale that no one has ever done before. However, the one thing that we can sort of mention as a negative from the collaboration with them is that the reason we hired them was that they were the experts in a never-have-been-done before the challenge. The fact that they kept sort of delivering late and asking for more time was something that made the production a bit challenging on our end sometimes — that's the only negative. 

Again, we didn't have any reservations in terms of the quality of what they delivered, and overall, we understood that it was a challenging topic so some delays were fine and acceptable — there were just a bit too many blanks.

How did Numlabs perform from a project management standpoint?

The model we had agreed on was actually on a time and material basis, and we agreed on a daily rate for their employees. This is what proved to be challenging on our side because, for both conceptual and personal reasons, work took longer than expected — this strained our budget a bit. Numlabs offered us though to fix some remaining issues without additional remuneration. The amount in the end, which was £30,000–£40,000, was more or less what we ended up paying them — but it was basically more than we had initially anticipated because the estimate of how much time they needed was not met in the end. 

They ended up having spent more time. Again, it's not necessarily all negative in that regard and we're generally happy with the collaboration. In terms of product management, we used Jira, which was a tool to assign tasks and organize the workflow. In general, our team works in a scrum methodology so we met every single day online and discussed the progress.

For communication basis, we used Slack for day-to-day communication, and then we used Google Hangouts or Google Meet for calls. They were responsive, nothing to complain about there.

What did you find most impressive about them?

I think they hold a truly unique position on the market, at least to our knowledge at the moment. Because first of all, they've got a relatively big team of machine learning engineers. Even though you can definitely find individual machine learning engineers in the market, it’s not very easy to find a whole team who is skilled in that space. Also, the other great benefit of Numlabs was that their team is quite versatile, I would say. Machine learning, even though it's quite specific already, it's still sort of divided into multiple different sub-disciplines. 

They provided the complete package that we needed and they were reasonably priced. The value they provided was quite high, so I think that makes their position in the market quite unique.

Are there any areas they could improve?

Just what I had said earlier about delays and meeting deadlines. That’s the only negative.

Any advice for potential customers?

I would definitely suggest having a technical project manager or someone who already is experienced to an extent with machine learning on the client's side — just to make sure that whoever speaks to Numlabs understands the technical language. It's not to say that Numlabs cannot communicate with a general product manager, but again, what we have experienced was the slight delays, so it needed a technical man on the client team.

That would be one bit of advice. The other one is probably quite general, not sure if it's related to Numlabs directly, but again, just always have a buffer for your projects when it comes to machine learning. It’s a pretty experimental discipline and one which can't be really predicted right away at the beginning of the production — just have some buffer in a contingency in your project.

Please describe your company and your position there.

I am co-founder and CTO of Createk. The company mission is to connect business and science via implementing cutting edge ideas and research results in industry. Our projects mainly include preparing, running, managing, and doing research in R&D projects for heavy industry companies like: steelworks, glassworks and foundries. The projects are usually worth about 5-30M PLN (about 1-6M euro) and last 1-2 years. Moreover we conduct our own R&D projects in 3D printing and scanning for the foundry industry.

For what projects/services did your company hire Numlabs?

We were looking for a partner to help us work with AI and big data technologies at all levels, starting from preparation of project plans through research up to implementation.

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

We found Numlabs thanks to networking and personal recommendations. After the first common project we decided to start full-time cooperation because of very good communication, business approach to solving problems, wide knowledge in AI and big data: timeseries, image and video analysis, data mining, predictive models, ETL and employees experience in the industry field.

Describe the project in detail and walk through the stages of the project.

Until this day, we prepared 6 R&D projects together, totaling more than 50M PLN (10M euro) and consulted the next few. Numlabs tasks were:  analysis of business needs and environment,  analysis of the state of the art based eg. on: research publications and industry news,  preparation of comprehensive research plan adjusted to bigger multidisciplinary projects.

Moreover, preparation of cost, risk and timeline analysis for research plans -exhaustive consulting meetings with stakeholders and other R&D stuff about prepared R&D plans.

How many resources from the Numlabs team worked with you, and what were their positions?

Usually we work directly with the CEO, CTO and R&D Lead. If needed Numlabs involved machine learning developers or even graphic designer when we needed present reports.

Can you share any outcomes from the project that demonstrate progress or success?

Numlabs provides high quality analysis and inputs for research plans - on time. They exhaustively explain alternatives, risks, business implications and answer our questions.

How effective was the workflow between your team and theirs?

We work mostly in an agile way, depending on the project type and phase. We communicate mainly via phone calls, Meets and emails. However sometimes we use chats like Mattermost. About project management tools, we use mostly Trello and Gitlab.

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

I feel that we work together in one team, not two separate companies, you can just call and ask about something not even connected with the project. What I really like is that they can quickly verify business problems in technology terms and if this is not their area of expertise tell us precisely, or even find, what kind of company can help us, better than Numlabs.

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

I think no, the cooperation is great so far.

Please describe your company and your position there.

SDA is an international company operating mainly in Central Europe. Provides IT and future skills training services to both individual and business clients. I am a business development manager in a B2B client department. My role is to ensure that the client's needs are met at the highest professional level.

For what projects/services did your company hire Numlabs?

The client has established a new department dedicated to AI. Our task was to train and re-train employees to successfully perform their new tasks. Numlabs was hired to develop and deliver a series of training courses on: data analysis, machine learning and data science

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

I met one of the founders during my post-graduate studies. I commissioned Numlabs to conduct demonstration lessons. Due to the good feedback from the client we continued to work together.

Describe the project in detail and walk through the stages of the project.

The first phase consisted of demonstration lectures. After these, the client decided to cooperate more extensively. In particular, to conduct training in the form of workshops. Numlabs not only prepared and conducted training, but also helped in determining customer needs. The scope of training included, among others: natural language processing, building neural networks, using cloud services in machine learning projects.

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

In addition to the founder with whom I had a business relationship, several senior machine learning and data science developers participated in the project. Their task was to prepare and conduct classes on the material in which they were experts.

Can you share any outcomes from the project that demonstrate progress or success?

I think the best indicator of success is that the client is satisfied and already extended our training

How effective was the workflow between your team and theirs?

Everything was fine, tasks were completed on time and communication was seamless

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

The Numlabs team really cared about delivering knowledge in a way that was accessible to learners.

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

I think the cooperation is exemplary