Numlabs - logo

How do we work

   

Development methodology of AI projects is a mixture of agile software building approach with meticulousness of scientific research.

AI in practice

Successful deployment of AI-enabled technology requires work in fields of R&D and software engineering, with expertise in machine learning and a deep understanding of a client's business.

The main difference from other software projects is the presence of a modelling step - an iterative process of Machine Learning model creation that consists of data analysis, feature engineering, testing and results analysis.

Final effect depends on input data quality, technology itself (model architecture) and precision in defining a business problem.

scrum image

Circle of AI project

AI project development cycle is the process of discovery - if and how a certain business problem could be tackled with the use of machine learning models. It requires both a lot of intuition and experience at the beginning and a scientific meticulousness and scrutiny in the process.

1. Needs analysis

01

  • identification of the project's business objectives

  • indication of recommended technical paths for project development

  • estimating costs and client engagement in the next phase of work

  • creating and discussing a report on the progress of the project

With a coherent vision of our common goals, we start cooperation.

2. Feasibility study

02

  • verification of the technical feasibility and market testing of the project

  • preparation of selected elements of the project for testing and demonstration purposes

  • identification of project risks

  • preparation of a project map and pricing

  • review of the feasibility report

We verified the project assumptions theoretically and experimentally. At this stage, the pricing, scope of work and timescale are more precise.

3. R&D i MVP

03

  • software development in 1-2 weekly cycles

  • meetings aimed at receiving feedback and directing changes

  • concise information on progress and obstacles in work

  • realisation of the main functionality of the project

  • testing in an environment similar to the production environment

4. Industrialisation

04

  • selection of the final hardware or cloud carrier

  • ensuring: scalability, reliability and security of AI project

  • integration with external systems

  • performing approval tests

At the end of this stage you get a complete product that successfully runs in a production environment.

5. Maintenance

05

  • monitoring and fixing of identified errors

  • counteracting bias and data drift

  • adapting the AI model to changes in systems

We will not leave you alone. You can count on our support both in project development and in fixing potential problems.

1. Needs analysis

01

  • identification of the project's business objectives

  • indication of recommended technical paths for project development

  • estimating costs and client engagement in the next phase of work

  • creating and discussing a report on the progress of the project

With a coherent vision of our common goals, we start cooperation.

explainability timeline throughput intuitiveness data availability performance latency budget usability

2. Feasibility study

02

  • verification of the technical feasibility and market testing of the project

  • preparation of selected elements of the project for testing and demonstration purposes

  • identification of project risks

  • preparation of a project map and pricing

  • review of the feasibility report

chart image lupe image

We verified the project assumptions theoretically and experimentally. At this stage, the pricing, scope of work and timescale are more precise.

3. R&D i MVP

03

  • software development in 1-2 weekly cycles

  • meetings aimed at receiving feedback and directing changes

  • concise information on progress and obstacles in work

  • realisation of the main functionality of the project

  • testing in an environment similar to the production environment

24h

Are you ready?

   

Sprint planning

   

defining experiment/ development scope for next week(s) and deliverables

Experimentation
/Development

   

according sprint plan

Daily scrum

   

everyday feedback and team coordination

Reporting
/Deployment

   

delivery of developed value

Demo

   

meeting with a client, presenting results and conclusions, feedback collection

Backlog refinement

   

upadting production vision and development way

Let's start another project!

   

4. Industrialisation

04

  • selection of the final hardware or cloud carrier

  • ensuring: scalability, reliability and security of AI project

  • integration with external systems

  • performing approval tests

  • przeprowadzenie testów akceptacyjnych

At the end of this stage you get a complete product that successfully runs in a production environment.

monitor image

5. Maintenance

05

  • monitoring and fixing of identified errors

  • counteracting bias and data drift

  • adapting the AI model to changes in systems

We will not leave you alone. You can count on our support both in project development and in fixing potential problems.

team image

Communication channels

Being a remote-first company we know, how critical is information flow. Here's what we use:

mattermost icon

Mattermost


A self-hosted company chat that ensures privacy.

cloud icon

Owncloud


To share the data with clients and between the team. It's encrypted and located on our private servers in EU.

gitlab icon

Gitlab


We use it in two ways - as a code versioning system and for tracking project level issues and tasks.

flags icon

Languages


Our development team is using polish and english. Our consultants also speak german, french and arabic.

Numlabs proved to us that thanks to their knowledge and experience, they can create solutions able to process a vast amount of data. The results of their work helped us strengthen our position on the market and impressed our customers.

— Andrzej Kwiecień, CEO

Apriside ApS

Numlabs have joined our project at a relatively early stage and have proven to be a professional and trustworthy partner until the end. Despite the challenging nature of the task, they managed to communicate each challenge and collaboratively work on solutions with the wider team. We will certainly work together again.

— Maciej Zasada, Technical Director

UNIT9 LTD

Working with specialists from Numlabs was a pleasure. The most important thing was a creative approach to both practical and scientific parts of the problem. The effect is outstanding, our support and resistance zones are one of its kind in the world and provides an edge to our users.

— Marcin Tuszkiewicz, CEO

Inwestio Sp. z o.o.