32 tools for developing AI solutions. The Ministry of Digital Transformation has published recommendations for AI startups

«The global market is experiencing a growing demand for responsible AI. Users of innovative products expect companies to provide breakthrough AI solutions that are also safe for them. These recommendations are a practical guide for Ukrainian developers on how to maintain a high level of innovation in products and at the same time design solutions while respecting human rights,» commented Oleksandr Bornyakov, Deputy Minister of Digital Transformation for IT Development, on the collection of recommendations.

The Ministry of Digital Transformation assures that the advice was developed based on the best global standards, in particular the HUDERIA methodology from the Council of Europe, which helps assess the impact of AI systems on human rights.

The 59-page document provides practical tips on how companies can choose a large language model and tailor it through training and fine-tuning. In particular, it recommends paying attention not only to the number of parameters or popularity of the model, but also to the quality of the training data, transparency, adaptability and reputation of the vendor.

A separate block contains tips on how to avoid copyright and property rights infringement, especially when using generative AI.

The application provides over 30 tools for developing AI solutions, detecting errors in code, auditing bias and fairness in systems, and more.

«Design privacy by default and design: minimize data collection, provide encryption, access control, logging, and the right to delete data. The user should control their information,» advises the Ministry of Digital Transformation.

The ministry believes that it is necessary to choose tools that focus on developing human-centered AI and that
designed with the rights and well-being of users in mind. Some platforms offer features that support ethical AI design, including tools for monitoring and assessing the impact of systems
AI per person.

What tools for developing AI solutions does the Ministry of Digital Transformation recommend?

1. Aequitas Bias and Fairness Audit Toolkit is an open-source toolkit for auditing the bias and fairness of machine learning models, developed by the University of Chicago.
2. IBM’s AI Fairness 360 is an open-source AI fairness toolkit designed to identify and eliminate biases in models.
3. Anthropic Claude is a generative artificial intelligence model used for text generation and natural language processing.
4. Apache Mahout is a scalable, open-source machine learning library designed for building recommendation, clustering, and classification algorithms.
5. AWS SageMaker Clarify is a tool for detecting biases and ensuring explainability of artificial intelligence, used to analyze models for biases.
6. Azure AI is a cloud-based artificial intelligence platform from Microsoft that provides services for developing, training, and deploying machine learning models.
7. Caffe is an open-source deep learning framework developed at the University of California, Berkeley, optimized for speed and modularity.
8. CNTK (Microsoft Cognitive Toolkit) is an open-source deep learning framework used to create neural networks.
9. Cursor is a code editor that uses machine learning technologies to provide intelligent suggestions, auto-completion, and error detection in code.
10. Deeplearning4j is a deep learning framework for Java and Scala with support for distributed computing.
11. Microsoft’s Fairlearn is an open-source toolkit for ensuring fairness in artificial intelligence, used to assess and reduce biases in models.
12. Fast.ai is a high-level deep learning library based on PyTorch, aimed at simplifying the development of machine learning models.
13. GitHub Copilot is an AI-powered tool developed by GitHub in collaboration with OpenAI. It integrates with code development environments like Visual Studio Code and provides recommendations and auto-completions as you write code.
14. Google Gemini is a generative artificial intelligence model used to automate tasks and create content.
15. Google Vertex AI is a cloud-based platform for machine learning and generative artificial intelligence, providing tools for developing and deploying models.
16. Google’s AI Explainability 360 is a set of tools for ensuring the explainability of artificial intelligence, used to analyze and explain model decisions.
17. H2O.ai is an open source machine learning platform used to build data analysis and prediction models.
18. Hugging Face Transformers is an open source library for natural language processing (NLP), providing pre-trained models and tools for working with text.
19. IBM Watson is an artificial intelligence platform that provides services for natural language processing, machine learning, and data analysis.
20. JAX is an open-source library for numerical computing from Google Research, used for developing deep learning algorithms and optimizing models.
21. Keras is a high-level neural network library that works with TensorFlow and is used to build machine learning models.
22. LightGBM is a high-performance gradient boosting framework from Microsoft, optimized for speed and performance.
23. Microsoft Copilot is an artificial intelligence tool powered by the Bing search engine and large language models from OpenAI, used to generate answers, search for information, and analyze data.
24. Midjourney is a generative artificial intelligence model that transforms text descriptions into images. It is used to create artistic and creative visualizations, providing the ability to generate images based on given text queries.
25. MXNet is an open source deep learning platform from Apache Software Foundation, supporting various programming languages.
26. ONNX Runtime is an open source tool for executing machine learning models in ONNX format, supporting various hardware platforms.
27. OpenAI ChatGPT is a generative natural language processing model from OpenAI, used to generate text and answer questions.
28. Perplexity is a web service from Perplexity AI, used for searching and processing information. PyTorch is an open source machine learning framework from Meta, used for developing and training models.
29. RapidMiner is a data analysis and machine learning platform that provides tools for building models without the need for programming.
30. Scikit learn is a Python library for machine learning, used for classification, regression, and clustering.
31. TensorFlow is an open-source machine learning platform from Google, used to create and train models.
32. Theano is an open-source Python library for numerical computation that allows you to efficiently define and optimize mathematical expressions.

Read the country’s main IT news in our Telegram

On the topic

Read the country’s main IT news in our Telegram

The Ministry of Digital Affairs has published instructions for booking employees through the «Diya» portal, which boils down to 4 steps

On the topic

The Ministry of Digital Affairs has published instructions for booking employees through the «Diya» portal, which boils down to 4 steps

PrivatBank launched an AI service for business that will instantly analyze the business reputation of counterparties

On the topic

PrivatBank launched an AI service for business that will instantly analyze the business reputation of counterparties

Search AI startup Perplexity releases tool that creates reports and spreadsheets in 10 minutes

On the topic

Search AI startup Perplexity releases tool that creates reports and spreadsheets in 10 minutes