Epinote

Please describe your company and your position there.

I work at Mobitouch as a COO. We design and scale products for enterprises, SMBs, and startups. We build teams composed of ambitions and talented software developers who deliver high-quality IT projects for global brands.

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

Mobitouch decided to cooperate with Epinote on a data cleansing project. As we have lots of data, a part of which was outdated, we needed to structure, update and recognize any conflicts, so we could make data-based decisions with more confidence.

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

We chose Epinote because we knew the COO and CEO, who we could trust with giving access to sensitive information. The pricing which they offered was also attractive in comparison to other companies providing data-related services.

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

The project consisted of a few iterations suggested by Epinote, so the whole process was very efficient and did not need rework. The scope of project included data validation, including indicating outdated or faulty records. Then the database was structured and standardized, missing data was added. As the last step, Epinote checked the whole dataset to get rid of duplicates and conflicts.

The technologies used for this project included SQL, Google Sheets and Python scripts.

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

Epinote’s COO, Kamil, was responsible for the project and handled communication with our team himself. We also got some valuable advice from Wojtek, their tech expert. They dedicated a project manager and a team of data analysts to work exclusively on our project.

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

As a result of this project we now have one consistent, correctly described and up to date database, which we use day to day running our firm and we can be confident making business decisions based on it.

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

The communication with Epinote is really something that sets them apart from most of the companies I ever worked with. We were using collaboration tools and digital communication channels. They always responded immediately and communicated in a clear, straightforward manner.

We were positively surprised by how few meetings enabled them to deeply understand our business and the project itself.

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

Epinote seems to be a really efficient organization. They knew how the processes should look like, but also tailored them to our requirements and preferred approach. I was impressed by how much initiative they showed.

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

Epinote delivered a great work, and I would say, exceeded expectations, so there is nothing I could suggest to improve the outcome. I strongly recommend them for data-related projects.

Please describe your company and your position there.

I am a Chief Data Scientist in ReSpo.Vision. In ReSpo, we aim to revolutionize sport analytics with use of the newest machine learning and deep learning models combined with mathematical modelling. As a Chief Data Scientist, my work revolves around solving the most difficult research problems that we encounter on our way.

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

Epinote has been providing data annotation services for a series of tasks connected with training deep learning computer vision models.

Data annotation projects are often much more difficult than their look may suggest. This is the case in ReSpo.Vision, as we have diverse annotation tasks, many of them significantly more complicated than basic annotation of visible objects.

Epinote excels in delivering high quality annotation, especially where the task at hand is not a trivial one.

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

We knew epinote’s co-founders and their drive to deliver high quality, no matter the task. Most important factor was the quality-to-price ratio, as epinote delivers quality of highest standard in the data annotation domain with prices significantly lower than big data annotation companies.

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

Series of data annotation projects were and are being realized in collaboration with epinote. Those were connected with training deep learning computer vision models.

Project starts with defining the annotation task - which classes of objects should be annotated, what kind of annotations will be used for each of them. Then the tutorials will be prepared and annotations tasks distributed between annotators. Afterwards, the datasets are being annotated. After initial annotation is done, all the images go through quality assurance to ensure that no mistakes are made.

The most important deliverable is the dataset itself, as it serves as a basis for deep learning models training.

In our collaboration, we use tools such as Pytorch, CVAT, OpenCV.

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

We have been working with three people from epinote’s team. Mateusz, CEO, with whom we have agreed on project details from the business angle. Kamil, COO, with whom we have fine-tuned technical project scope and deliverables schedule. Krzysztof, Project Manager, who assured that all deliverables were delivered on time and with high quality ensured.

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

Every project results in a dataset or a set of datasets annotated according to the task definition. Those datasets are then used to train machine/deep learning models, so the data quality must be impeccable or the models will be fed with incorrectly annotated objects, which will impair their accuracy. This makes the annotation quality itself the key metric for project success.

In ReSpo.Vision, there are many difficult (much more difficult than the “industry standard”) and diverse annotation tasks with a notable number of edge cases. Where epinote truly excels is the ability to ensure that those difficult datasets are annotated both quickly and faultlessly.

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

Epinote is very agile. They are able to scale up and down efficiently to adjust to data annotation volume which is required in a given moment.

Every project starts with a series of calls with their management team, so both sides can agree on both the business (timeline, rates, method of delivery) aspect and achieve consensus on the technical details of the task.

Afterwards, most of the communication is done via Slack, whenever there is an update from their side or a need from our side to contribute on how to approach a particular difficult annotation case.

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

Epinote's team understands exactly how data annotation fits into the machine learning projects and models lifetime, therefore they know how to deliver high quality data which will result in accurate models (pointing out edge cases, working with the client to correctly understand the task). They are also able to quickly recruit and set up a larger number of annotators to deliver the required datasets.

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

There were no issues in collaboration with epinote, therefore we can recommend them with highest confidence.