Machine learning & artificial intelligence
Over the past few years, there has been a real revolution in both software and hardware, which has given us many new opportunities to use smart solutions in more and more industries and products.
Many companies have already jumped on board and taken advantage of this, while others are just beginning to catch on. No matter where you stand, you can always count on Neodev as a skilled partner to collaborate with to achieve your goals.
-
Python
-
U2-net
-
pytorch
-
GCP
-
CNN
Customer case
Boni AB
Take a look at the image next to this text. Do you see a bird? Do you see a bird standing on a wall? Do you see a bird standing on a wall and looking out over a harbor?
Object detection
Even if one option is claimed, the other observations are also correct. Many would probably first say that it is a bird (or perhaps even a seagull?) in the image. These people can likely easily draw a line around the object to indicate which pixels constitute the bird and which are something else.
It may seem obvious, but when letting a computer figure out which part of the image is of interest, it becomes a much more difficult problem. Shouldn’t the boats in the background be equally prominent objects as the bird?
Boni
Boni receives a large number of images showing products with various backgro…
Experts on many levels.
Natural language processing.
Tasks involving language understanding have long been the domain of humans. But thanks to machine learning, we can now let computers do the heavy lifting for us! NLP technology enables, for example, automatic speech recognition and document sorting entirely without human interaction.
[Linguistic processing]
[Feature detection]
Voice identification.
Voice identification technology has advanced significantly in recent years and can now be used to recognize different speakers with high accuracy. This is useful for everything from security measures to customer service and healthcare. The technology can help streamline processes and improve the customer experience
Computer vision.
Valuable information can often be found in collected images and videos. With machine learning, such data can be extracted - entirely without human involvement. Tools from this field have recently been successfully used in areas such as optical diagnosis or object localization in collected videos.
Time series.
Time series is a large subfield within sequential data, consisting of data where each measurement point also has a time stamp. Classic examples are different stock indices, weather and sensor data. In the business world, typical use cases for AI/ML include forecasting a time series’ future development or classifying a (or a part of a) time series, for example, to detect changes in behavior (also see predictive maintenance).
Predictive maintenance.
The fourth industrial revolution opens up a myriad of opportunities for improvement in the business world. One such opportunity is predictive maintenance, the ability to predict remaining lifespan of equipment based on current measurements. Machine learning allows us to minimize time losses due to equipment failures and reduce unnecessary resource usage.
[Time to be series]
We teach you more.
[Seminars]
[Courses]
[Workshops]
Neodev regularly organizes interesting events within the field of machine learning and AI, where one of our experts shares their knowledge.
Get in touch if you or your company want to learn more!
How we work with machine learning.
Identification and selection of use case
- Workshop (if necessary) to identify use cases for AI within their business. If proposals already exist, this step can be skipped.
- Evaluate proposed AI use cases using our evaluation process. We try to estimate whether the proposed use case is meaningful (i.e., improves profitability, equality, or climate impact) and ensure that the prerequisites for successful implementation are met.
- Find “minimum AI model performance” requirements.
Technical requirements specification
- What type of platform, e.g., embedded, cloud, on-premise?
- Is there a need for a user interface?
- What are the performance requirements, such as response time?
- What is needed in terms of data infrastructure, such as databases and data pipelines?
- What is the need for continuous training, evaluation, and deployment?
Project planning
- Critical points
- Estimate time required
Project start
- Continuous contact and follow-up according to an agile working method (scrum/kanban)
- Ongoing reporting during implementation
Delivery and final reporting
Collaboration
AI for real.
AI Sweden is the Swedish national center for applied artificial intelligence, consisting of more than 100 partners within the business sector, academia, and the public sector.
As a partner of AI Sweden, we are part of their mission to accelerate the use of AI in Sweden for the benefit of our society, competitiveness, and for everyone who lives here. Thanks to this partnership we have access to new tools and innovative technology, knowledge, and a network of 100+ partners from various sectors in Sweden.
Our contribution is to share our experiences, both technical and business-related!
Some previous projects.
Home alarm
The overarching problem is to detect intruders while avoiding false alarms from other sources, such as pets.
The work involves developing various types of classification models for time-series data and image data.
In addition to model development, requirements gathering, data collection and annotation, development of training frameworks on cloud services, and more, will be implemented.
Classification system
AI module for integration into an existing system to categorize textual data into various classification systems. The work included text pre-processing and implementation of models for word embedding.
Frontend development for the same system - developing a user interface in React to handle input data, demonstrate categorization, and facilitate data collection.
The same client chose us to modernize their existing web application, which was written in jQuery and web forms. The development was done in React, and the work required planning, restructuring, and development of a new workflow to deliver a cloned version of the web application with additional features.
Sample gathering
Development and design of a “mobile-first sample gathering” web application with a serverless backend in Azure. Implementation of model training and versioning pipeline. The models were then used for real-time prediction of another web application client using Tensorflow JS.
Preliminary study
Preliminary study to investigate the possibility of applying AI, data science, and machine learning methods in a specific area for a customer in the mobile industry.
Embedded system
Software development in Machine Learning for use in embedded systems for a customer in the security industry. Mainly using C and C++ as programming languages, and Python and shell scripts for developing testing and training frameworks, as well as data analysis.