I am ready for a long road flight for work with a week- or months-long projects.
Cookies managing
Cookie Settings
Cookies necessary for the correct operation of the site are always enabled.Other cookies are configurable.
Elevate Your ML Career with Practical Generative AI
Say goodbye to shallow tutorials! Learn from industry experts and develop practical expertise in Generative AI.
Machine learning engineers and data scientists
Software engineers and developers
ML professionals
who are looking to master LLMs and Generative AI
who want to move into AI/ML Engineer roles
who look to integrate and scale LLM-driven applications, features, or products
Best Suited For
Topic of the lecture:
Share
Alexey
Niv
Ekaterina
Course starts on: October 28th
Why Choose This Course?
Learn from GenAI engineers and researchers.
Insights from the experts
Get a deep understanding of what is under the hood of GenAI models.
Deep dive into generative AI
Master both fundamentals and cutting-edge AI trends.
Future-proof your skills
1
2
3
Stay informed on the most recent advancements and research in Generative AI.
Engage with the latest research
Learn how to use multimodal LLMs and how to generate, personalize, and edit images.
Multimodal skillset
Learn not only how the LLMs work, but also how to deploy them and how to boost their inference efficiency.
Work with LLMs in production
4
5
6
Entry Requirements
We expect that our students are familiar with ML basics (training and evaluation of models, loss functions and metrics, hyperparameters) and have some experience with neural networks (at least fully connected and convolutional).
Machine learning &
deep learning basics
During the course, the practice and home assignments will be in Python. We expect candidates to have practical Python skills before starting the course.
Coding skills
AI Applications requires high school math while Creation needs a bachelor’s STEM degree with university-level math proficiency. We won’t teach topics from scratch but offer a quick recap before advanced topics.
Statistics & math
Curriculum Designed for Professionals
Free
Part 1: Intro to LLMs
Part 1: Intro to LLMs
- What are LLMs and what are they capable of.
- LLM training stages: pre-training, supervised fine-tuning, preference training.
- LLM APIs. How to choose and how to use them.
- Deploying an LLM API-based chat service.
- Making the most of an LLM API: prompt engineering, few shot learning, chain-of-thought, chaining etc.
- Retrieval Augmented Generation (RAG).
- LLM agents.
- Creating a RAG system.
- Hugging face ecosystem and open source LLM zoo.
- LLM scaling laws.
- Deploying an open source LLM; LLM inference engines.
Part 1: Understanding language modeling
- Language modeling.
- RNN-LM.
- Sampling strategies.
- Seq2seq with Machine Translation as an example.
- Attention mechanism.
- Self-attention and the original transformer architecture.
- Encoder-only models.
- Development of the self-attention mechanism.
- Positional encoding.
- Long-context adaptations.
- Mixture-of-Experts.
- Math of RLHF and DPO.
- Less is more in alignment.
- RNN-based and hybrid architectures. Griffin and xLSTM.
- State space models and hybrid architectures. Mamba, Jamba, and Samba.
- Efficient attention.
- KV-cache and its optimizations.
- Speculative decoding.
- Core concepts of working with GPUs. GPU memory structure.
- Consumer GPUs.
- Introduction to distributed training. Process-based communication.
- Parameter Server architecture.
- Data-parallel training of neural networks.
- Model parallelism, gradient checkpointing, offloading, sharding.
Part 1: Transformers outside NLP
- ViT.
- CLIP.
- AE, VAE & VQ-VAE.
- Math of VAE loss.
- GANs.
- Inversion and latent manipulations with GANs.
- VQGAN.
- Forward & backward process, latent diffusion, conditional diffusion.
- Math of DDPM. SDE and continuous-time diffusion.
- Stable diffusion and sd-turbo.
- Guidance: classifier, classifier free.
- Editing by feature manipulation. Prompt-to-prompt, Plug and Play.
- Editing by fine-tuning. InstructPix2Pix, Text Inversion, ControlNet.
Part 4: Customization
- Customization vs RAG.
- Parameter-efficient fine tuning, LoRA, QLoRA.
- Best practices of model improvement.
- Multimodal LLM (MLLM) architectures.
- Deploying MLLMs
- Performance risks: relevance, hallucinations, user interaction.
- Ethical risks: toxicity, bias.
- Safety risks: jailbreaking, data poisoning, prompt injection.
- Existential threats of AI and AI regulations.
- Artificial data, risks and detection.
- Pre-deployment LLM evaluation.
- Evaluation of a deployed LLM. Tracing, LLM observability. Regression testing. Experiment management. LLM monitoring. Explicit/implicit feedback.
- Particular metrics: from simple sanity checks to LLM as a judge.
- Inference-time quantization.
- Multi-GPU deployment.
Meet Your Team of Industry Experts
Noa Lubin
Niv Haim
Stanislav Fedotov
Alexey Bukhtiyarov
Director of Data Science at Fido
ML researcher at Weizmann Institute of Science
AI content lead at School of AI & DT
NLP Team Lead at Ex-human
Noa is a Data Scientist, working as Director of Data Science at Fido. Formerly worked as a researcher at NASA, at Diagnostic Robotics, Amazon, Elbit and IAI.
Niv Haim is a computer vision and Machine Learning researcher at Weizmann, where he earned his PhD under the guidance of Prof. Michal Irani.
With a rich background in arranging educational programs in the field of Data Science, Stanislav is deeply passionate about AI education.
Having a vast experience with applied DL and LLMs, Alexey is leading an NLP team at AI startup. His background also includes distributed training research and Multi-GPU programming.
Entry Process
Admissions are closed once the requisite number of participants enroll for the upcoming cohort. Apply early to secure your seat.
1
2
3
Application
Free part
Payment & enrolment
2 weeks
Submit your application by filling out the form. You will receive an email with access to the free part of the program and the following stages of the process.
The free part will take ~ 2 weeks. You need to complete the free part and the assignments before the cohort begins.
Last call for starting: October 14th
If you successfully pass the free part, you'll proceed to payment and enrolment into the program.
Cohort starts: October 28th
The free part must be completed by November 1st or you will only be able to get in the next enrolment
Built for Working Professionals
The time required for studying materials, completing homework, and attending consultations usually takes up to 10 hours per week.
New study materials, lectures, and homework assignments to review
Deadline for submitting homework from the previous week
Q&A session with experts and instructors
Monday
Wednesday
Once a week
What Our Alumni Are Saying
Alexander Kazakov
OCR
Computer Vision
Machine Learning
Hello, my name is Alexander Kazakov, and I work with OCR and text extraction from images and PDF files at Megaputer Intelligence, a company specializing in text data analysis...
Hello, my name is Alexander Kazakov, and I work with OCR and text extraction from images and PDF files at Megaputer Intelligence, a company specializing in text data analysis.
My decision to study in the Generative AI program stemmed from a desire to learn something new...
My decision to study in the Generative AI program stemmed from a desire to learn something new...
Read more
ML
Andi Mardinsyah
Data Scientist
Hello! My name is Andi Mardinsyah, and I work as
a Data Scientist at Telekomunikasi Indonesia.
My company is currently working on creating applications for natural language processing...
a Data Scientist at Telekomunikasi Indonesia.
My company is currently working on creating applications for natural language processing...
Read more
Emanuele Bezzecchi
AI Roadmap Manager
The videos taught me a lot and give a better understanding of LLMs, really helpful in my job I underestimate the ratio between free time/time needed to do the homework.
I like to really understand what I’m doing and so I do not finish 3 of the 5 homework...
I like to really understand what I’m doing and so I do not finish 3 of the 5 homework...
Read more
Ahmad Zeidan
Developer Support Engineer
The course over all is great I'm enjoying it so far, the videos are vary good and easy to understand, long format reading and papers are ok as well, I'm kinda used to reading similar things in university, and have gpt-4 to help 🙂
Emanuele Antonioni
Machine Learning Engineer
Until now I am finding the course great! I really enjoyed the first two classes, the third was a bit less practical but still really interesting. The homeworks are really good, maybe sometimes a bit too long, but really enjoyable. Until now my feedback is totally positive.
Igor Samenko
DS, ML & DL Engineer
The course is great! I really like it!I like the amount of new material and the number of articles (referenced in the lectures). I believe that the knowledge gained in this course will be highly relevant for the next few years. Personally, I like the more technical...
Read more
Semyon Abramov
Machine Learning Engineer
I'm excited to announce that I've completed the Practical Generative AI course offered by School of AI and Data Technologies. Over the span of four months, this program provided me with hands-on experience in various topics such as...
Read more
Get Started with Free Generative AI Basics. Decide on Your Next Step Later
Up-front
Everything you need to start working with generative ai
$1,500
$500 for each of the three modules
Per module
Split your payment into affordable chunks and pay every module
$500
You will be able to purchase the Module 2 after completing the Module 1
AI Insights
FAQ
Applications & admission process
The admission process consists of two steps.
First, candidates submit an application on the website, provide a link to their LinkedIn profile or resume, and gain access to the free introductory part of the course. During the introductory part, the candidate watches lectures, studies long reads, and completes several auto-graded assignments.
Upon completing the free introductory part, the candidate can pay for the first module or the entire program and join the next cohort.
First, candidates submit an application on the website, provide a link to their LinkedIn profile or resume, and gain access to the free introductory part of the course. During the introductory part, the candidate watches lectures, studies long reads, and completes several auto-graded assignments.
Upon completing the free introductory part, the candidate can pay for the first module or the entire program and join the next cohort.
AI&DT School is dedicated to creating a motivating and supportive learning environment. To ensure success, we need to confirm that our candidates have the availability and capability to succeed in the intense study program we offer.
You are expected to have a solid grasp of Python programming or another scripting language. Prior experience in basic/classical ML modeling or prototyping will be of immense benefit when doing this course. You should also have beginner-level comfort with Deep Learning concepts.
The first Generative AI module requires only high school-level math.
In the second and third Generative AI Creation modules, we assume that our students have at least a bachelor STEM degree or its equivalent. So, we expect proficiency in first-year university-level math. As a preparation for this module, we suggest completing Mathematics for Machine Learning specialization on Coursera or a similar level course.
In the second and third Generative AI Creation modules, we assume that our students have at least a bachelor STEM degree or its equivalent. So, we expect proficiency in first-year university-level math. As a preparation for this module, we suggest completing Mathematics for Machine Learning specialization on Coursera or a similar level course.
The first Generative AI module requires only high school-level math.
In the second and third Generative AI Creation modules, we assume our students have at least a bachelor STEM degree or its equivalent. So, we expect proficiency in first-year university-level math. During the course we won’t teach these topics from scratch, but we will provide a quick recap of several advanced topics before they are required at the course.
In the second and third Generative AI Creation modules, we assume our students have at least a bachelor STEM degree or its equivalent. So, we expect proficiency in first-year university-level math. During the course we won’t teach these topics from scratch, but we will provide a quick recap of several advanced topics before they are required at the course.
During the course, the practice and home assignments will be in Python. We expect candidates to have practical Python skills before starting the course.
Program information
The next cohort will launch on October 28th, 2024.
You must complete the open part of the course before starting the cohort. This will take about 2 weeks. We will open access to the free part of the course before the cohort begins so that you have time to complete it.
Lectures are pre-recorded and can be viewed anytime. However, each week a home task is given with a one-and-a-half-week deadline.
You must complete the open part of the course before starting the cohort. This will take about 2 weeks. We will open access to the free part of the course before the cohort begins so that you have time to complete it.
Lectures are pre-recorded and can be viewed anytime. However, each week a home task is given with a one-and-a-half-week deadline.
The free part will cover the LLM basics and RAG. You will learn how to create services based on LLM APIs and how to deploy open source LLM into cloud.
Generative AI is at the forefront of the AI revolution. Our 360° curriculum on GenAI is designed to equip you with a deep understanding of LLMs, Multimodal LLMs, and Diffusion Models. You will get both a deep understanding of internals and a practical experience of deployment, fine-tuning, monitoring AI models. The field of Generative AI develops rapidly, but we constantly update the course and teach state-of-the-art concepts and technologies. And all along the way, our team of experts will be here for you to advise you and share their expertise.
This is not a short and cute recorded videos-based course – it is a professional upskilling program taught by industry practitioners meant to give your ML expertise more depth and variety.
This is not a short and cute recorded videos-based course – it is a professional upskilling program taught by industry practitioners meant to give your ML expertise more depth and variety.
Generative AI is becoming increasingly important in today's technological landscape. Its influence is growing across various sectors, bringing new approaches to content creation, data analysis, and problem-solving. Industries such as entertainment, software engineering, healthcare, and finance are seeing significant advancements due to this technology.
For professionals and researchers, a thorough understanding of Generative AI offers valuable opportunities. It allows them to work with state-of-the-art tools and contribute to innovative projects. As the job market evolves, expertise in this area can provide a competitive edge.
For professionals and researchers, a thorough understanding of Generative AI offers valuable opportunities. It allows them to work with state-of-the-art tools and contribute to innovative projects. As the job market evolves, expertise in this area can provide a competitive edge.
The Advanced Generative AI Program comprises live recorded videos and long reads. There are also weekly live Q&A sessions and regular live paperwatch sessions. While we encourage our students to attend, we also share recordings of all the sessions.
Yes. You will be able to access all class recordings even after you complete the program. As an alumni, you will also be invited to the alumni community where we share new materials and insights on AI development.
The projects in the Generative AI Program will include fine-tuning LLMs, deploying them and scaling these to a bigger industry use cases along with multimodality aspects. The goal of these projects would be to stay in sync with the latest advancements in the field of generative AI and its in-depth application.
Yes, you get a certificate of completion. To get the certificate you should complete all the required coursework - at least 80% of homework assignments submitted and graded as passed (additional module-level requirements may apply and will be announced at the beginning of the module).
Time commitment
The weekly workload can be up to 10 hours with 2-3 hours of videos or supplementary materials reading and extensive homework.
Cost & payment options
The full program cost is $1,500.
You can also purchase Generative AI by modules. Each module costs $500. Each subsequent module can be purchased after completing the previous one.
You can also purchase Generative AI by modules. Each module costs $500. Each subsequent module can be purchased after completing the previous one.
Sure. Most students have their tuition paid for by their employer. We can provide the course syllabus, official acceptance letter and other documents after the student has completed the free part of the course.
Powered by Nebius
Nebius AI, established in late 2023, is a leading AI-centric public cloud platform designed to support the entire machine learning lifecycle. With a focus on empowering ML practitioners, Nebius offers comprehensive infrastructure and aims to become the preferred platform for generative AI developers.
Visit Nebius
School of AI and Data Technologies, 2024. All Rights Reserved ®
Manage Cookie Policy
Powered by TripleTen
Alexander Kazakov
OCR
Computer Vision
Machine Learning
Hello, my name is Alexander Kazakov, and I work with OCR and text extraction from images and PDF files at Megaputer Intelligence, a company specializing in text data analysis.
My decision to study in the Generative AI program stemmed from a desire to learn something new. This choice was driven by my aim to keep up with modern technologies and the latest developments. In my work, I already have experience in training neural networks, but in a different area and on a smaller scale.
The module proved to be very informative, and the provided educational materials were relevant to me. I'd like to note the availability of experts for discussions; they respond quickly and comprehensively to questions. On average, I spent 15 to 20 hours per week on the training, but I believe the pace of learning is individual.
Particularly memorable was the first week of the program, an introduction to LLMs led by lecturer Yuval Belfer. The new information and understanding of previously complex concepts as simpler opened new horizons for thinking and action.
The practical assignments were also interesting, especially the last project on information extraction from databases.
The only aspect I'd like to see improved is a deeper exploration of theory. I've already provided this feedback to the team, and they explained that this is a characteristic of the first module. The second module promises a more extensive study of theory. We'll see.
I like the program and trust its creators. I would recommend it to my colleagues involved in machine learning and text analysis. I think the first module might seem less interesting to them due to their existing knowledge, but the second module will be beneficial for a deeper dive into theory and the mechanics of model operation. For those familiar with programming but new to neural networks, the first module will be particularly interesting.
My decision to study in the Generative AI program stemmed from a desire to learn something new. This choice was driven by my aim to keep up with modern technologies and the latest developments. In my work, I already have experience in training neural networks, but in a different area and on a smaller scale.
The module proved to be very informative, and the provided educational materials were relevant to me. I'd like to note the availability of experts for discussions; they respond quickly and comprehensively to questions. On average, I spent 15 to 20 hours per week on the training, but I believe the pace of learning is individual.
Particularly memorable was the first week of the program, an introduction to LLMs led by lecturer Yuval Belfer. The new information and understanding of previously complex concepts as simpler opened new horizons for thinking and action.
The practical assignments were also interesting, especially the last project on information extraction from databases.
The only aspect I'd like to see improved is a deeper exploration of theory. I've already provided this feedback to the team, and they explained that this is a characteristic of the first module. The second module promises a more extensive study of theory. We'll see.
I like the program and trust its creators. I would recommend it to my colleagues involved in machine learning and text analysis. I think the first module might seem less interesting to them due to their existing knowledge, but the second module will be beneficial for a deeper dive into theory and the mechanics of model operation. For those familiar with programming but new to neural networks, the first module will be particularly interesting.
ML
Andi Mardinsyah
Data Scientist
Hello! My name is Andi Mardinsyah, and I work as a Data Scientist at Telekomunikasi Indonesia.
My company is currently working on creating applications for natural language processing, such as segment analysis, and also on projects related to LLM. That's why I decided to study in the Generative AI program – to understand this topic deeper and solve work tasks more effectively.
I had to choose between two Generative AI programs, but I chose the program from School of AI and DT because I really liked its curriculum. As you know, generative AI is everywhere now, and the technologies are developing very fast. It's hard to find an educational program that combines both theory and practice. In my opinion, the curriculum of this program is very complete and comprehensive.
I have finished the first module of the program, which is dedicated to Generative AI applications. I really liked this module and found it extremely useful. Although I am a data analysis specialist and new to Generative AI, I can confidently say that my time was well spent. Especially valuable was the fact that we did a lot of coding during the training, which is an important part of the educational process.
Besides the program content, I would like to highlight its organization. I have a busy work schedule and doubted if I could combine work and study. I assumed the lectures would be long, but was pleasantly surprised to find out that the recorded lectures last only 15 minutes and cover a lot of material. This allows me to spend more time on practical tasks, which are plentiful in the program. It's important to note here that this is not just one 15-minute lecture per week, there are usually several.
I will definitely recommend this program to my colleagues.
I think these materials will be useful for all Data Scientists who are involved in natural language processing and LLM.
My company is currently working on creating applications for natural language processing, such as segment analysis, and also on projects related to LLM. That's why I decided to study in the Generative AI program – to understand this topic deeper and solve work tasks more effectively.
I had to choose between two Generative AI programs, but I chose the program from School of AI and DT because I really liked its curriculum. As you know, generative AI is everywhere now, and the technologies are developing very fast. It's hard to find an educational program that combines both theory and practice. In my opinion, the curriculum of this program is very complete and comprehensive.
I have finished the first module of the program, which is dedicated to Generative AI applications. I really liked this module and found it extremely useful. Although I am a data analysis specialist and new to Generative AI, I can confidently say that my time was well spent. Especially valuable was the fact that we did a lot of coding during the training, which is an important part of the educational process.
Besides the program content, I would like to highlight its organization. I have a busy work schedule and doubted if I could combine work and study. I assumed the lectures would be long, but was pleasantly surprised to find out that the recorded lectures last only 15 minutes and cover a lot of material. This allows me to spend more time on practical tasks, which are plentiful in the program. It's important to note here that this is not just one 15-minute lecture per week, there are usually several.
I will definitely recommend this program to my colleagues.
I think these materials will be useful for all Data Scientists who are involved in natural language processing and LLM.
Emanuele Bezzecchi
AI Roadmap Manager
The videos taught me a lot and give a better understanding of LLMs, really helpful in my job I underestimate the ratio between free time/time needed to do the homework.
I like to really understand what I’m doing and so I do not finish 3 of the 5 homework. As example here is 7:38 in the morning and 7-8 in the morning is the only slot available in these weeks for me to follow lessons/do homework.
Anyway now that I get the way you teach I can honestly say that the technical content is good and I think to have spent my money in a good way. That’s my feeling.
I like to really understand what I’m doing and so I do not finish 3 of the 5 homework. As example here is 7:38 in the morning and 7-8 in the morning is the only slot available in these weeks for me to follow lessons/do homework.
Anyway now that I get the way you teach I can honestly say that the technical content is good and I think to have spent my money in a good way. That’s my feeling.
Ahmad Zeidan
Developer Support Engineer
The course over all is great I'm enjoying it so far, the videos are vary good and easy to understand, long format reading and papers are ok as well, I'm kinda used to reading similar things in university, and have gpt-4 to help 🙂
Emanuele Antonioni
Machine Learning Engineer
Until now I am finding the course great! I really enjoyed the first two classes, the third was a bit less practical but still really interesting. The homeworks are really good, maybe sometimes a bit too long, but really enjoyable. Until now my feedback is totally positive.
Igor Samenko
DS, ML & DL Engineer
The course is great! I really like it!I like the amount of new material and the number of articles (referenced in the lectures). I believe that the knowledge gained in this course will be highly relevant for the next few years. Personally, I like the more technical (theoretical) dive into technology and into math but I realize the course has a different format and that's fine with me.
The teaching team is wonderful. Quality of lectures and presented material 10/10. I like the lecturers and how they present the material. I see passionate people who love what they do.
I like the course syllabus and that the course gives a wide overview of many areas. But, the topics "Bias in Generative AI" and "AI safety" are currently the most controversial for me. I mean, yeah, it's "good to know" information. But it's not deep enough for me to be useful or something I could apply to my work/life.
The material in the long reads is well prepared, compressed and interesting. I like that the lectures don't give 100% on the answers in quize and you have to work to get the knowledge for the right answer.
I also like that the homework is based on new, actual technology. Also the "paperwatch" channel is a treasure trove of recent hot topics. Really love it ❤️ I hope to be able to keep access to this channel after the course finishes.
The teaching team is wonderful. Quality of lectures and presented material 10/10. I like the lecturers and how they present the material. I see passionate people who love what they do.
I like the course syllabus and that the course gives a wide overview of many areas. But, the topics "Bias in Generative AI" and "AI safety" are currently the most controversial for me. I mean, yeah, it's "good to know" information. But it's not deep enough for me to be useful or something I could apply to my work/life.
The material in the long reads is well prepared, compressed and interesting. I like that the lectures don't give 100% on the answers in quize and you have to work to get the knowledge for the right answer.
I also like that the homework is based on new, actual technology. Also the "paperwatch" channel is a treasure trove of recent hot topics. Really love it ❤️ I hope to be able to keep access to this channel after the course finishes.
Semyon Abramov
Machine Learning Engineer
I'm excited to announce that I've completed the Practical Generative AI course offered by School of AI and Data Technologies. Over the span of four months, this program provided me with hands-on experience in various topics such as
- Large Language Models (LLMs’ architecture and training, RAG, RLHF, DPO)
- Transformers in vision (ViT, CLIP, BLIP, Flamingo)
- Generative Models (VAEs, GANs, Diffusion Models, Controllable Generation)
- Efficient Deep Learning (distributed training, quantization)