The AI and Digital Transformation Podcast

We’ve covered episodes about AI in logistics in the past. Let’s now focus our attention to manufacturing. Some AI applications in this sector include predictive maintenance, quality assurance using computer vision, anomaly detection, and digital twins.  Building these solutions takes time and requires accuracy. If developed and operated manually 100%, this approach risks making more errors in your ML models and datasets. Rather than relying on grit, we can automate the entire process and let the AI solution run like clockwork. MLOps (machine learning operations) automates the ML process together with the help of a feedback loop system. It can be divided into layers, like a layered cake. Some of its layers include experiment tracking, dataset monitoring, qualitative tests and explainability layer.  In this episode, we talk to Marek Tatara, Chief Scientific Officer of DAC.digital as he tells us more about how MLOps works, and their experience in building a customized MLOps for the semiconductor industry under a large cooperative EU-funded project.  Listen to our episode if you want to make your ML project more effortless and more reliable at a larger scale.  Who is Marek Tatara? Marek Tatara, PhD - Chief Scientific Officer and Tech Lead of the AI team at DAC.digital, Assistant Professor at Gdańsk University of Technology, AI/ML Expert at M5 Technology, Member of the Polish Society For Measurement, Automatic Control And Robotics. At DAC.digital, he works on the research agenda of the company and on the implementation of both EU-funded and commercial R&D projects from the field of Computer Vision (especially multi-camera setup, 3D reconstruction, and object detection and tracking), Machine Learning (mainly for computer vision, e.g., object detection, DNN optimization, semantic segmentation), and Embedded Systems (wireless communication for IoT devices and medical devices implementation. Where to find Marek: DAC.Digital website LinkedIn Resources: Book recommendation: Modern Control Theory by William Brogan Aims50 - Artificial Intelligence in Manufacturing leading to Sustainability and Industry 5.0 Time Stamps (00:00:00) Trailer (00:01:08) Who is Marek Tatara? (00:01:50) AI in Manufacturing: Applications (00:04:18) Concepts behind MLOps (00:08:18) Custom vs pre-made tools for MLOps (00:10:41) Building an MLOps project is like building a layered cake (00:17:08) Adopting MLOps among small and medium manufacturing companies (00:19:26) Working in an EU funded ML project (00:21:06) Advice on AI adoption and implementation for SMEs(00:26:11) Final remarks and book recommendations --- More on G.M.S.C. Consulting Follow us on our socials: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠LinkedIn⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠YouTube⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Book an appointment⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ with us. ⁠⁠⁠⁠⁠⁠⁠⁠Sign up to our newsletter⁠⁠⁠⁠⁠⁠⁠⁠. --- Music credits: storyblocks.com Logo credits: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Joshua Coleman, Unsplash⁠⁠⁠

Does your business need an AI computer vision co-pilot? You might need it, especially if you’re working on a high-stake AI project that requires precision or accuracy. In this episode, Akridata’s AI engineer Alexander Berkovich tells us more about it as he covers the different use cases that have used Akridata’s computer vision co-pilot. To name a few are corrosion detection, autonomous vehicles, railroad inspection, and industrial maintenance. You can learn more about good data management and deployment practices to ensure a less biased operating AI model for your computer vision project. Who is Alexander Berkovich? Alexander is a principal AI/ML engineer at Akridata, whose tools and services save time and lower costs developing vision-based applications and systems. Previous positions include an R&D manager, team lead, and algorithm developer in a variety of domains, ranging from smart cities, to medical quality inspections, manufacturing and more, all in the computer vision space. His aim is to automate decision making based on a combination of visual sensors, software, hardware and the maths behind it all, to improve the quality of services, products and daily life. In addition to focusing on the technical aspects of development, Alex advocates for the importance of grasping the business case and employing high-quality data, especially in this AI driven era. Where to find Alexander: Akridata.ai LinkedIn Time Stamps (00:00:00) Trailer (00:01:42) About Alexander and Akridata (00:03:42) About computer vision copilots (00:07:15) Use cases (00:21:50) Importance of data quality when training models (00:16:01) Model training and deployment, accuracy, precision and recall (00:20:58) Dealing with clients’ needs (00:24:44) Addressing biases in AI computer vision models (00:47:44) Closing remarks  --- More on G.M.S.C. Consulting Follow us on our socials: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠LinkedIn⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠YouTube⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Book an appointment⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ with us. ⁠⁠⁠⁠⁠⁠⁠Sign up to our newsletter⁠⁠⁠⁠⁠⁠⁠. --- Music credits: storyblocks.com Logo credits: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Joshua Coleman, Unsplash⁠⁠⁠

Do you want to build the ultimate search engine that works perfectly for your business? A multimodal system processes all types of information like images, videos, audio, and text. It can end up either as a Frankenstein or a Swiss knife depending on how clunky or how smoothly these models communicate with one another. In this episode of the AI and Digital Transformation Podcast by G.M.S.C. Consulting, we had a chat with Duarte Carmo about building a customized multimodal search engine. We learned about his experience on building a multimodal search engine for a startup, and things you need to consider when building an AI product that does not rely heavily on big tech’s costly services.  If you like tinkering and using non-conventional methods to build personalized AI products like a multimodal search engine, this is the episode for you.  Who is Duarte Carmo? Duarte is a Portuguese technologist who is now based in Copenhagen. He loves finding ways on making technology improve people’s lives. He solves problems end-to-end by combining his interests and work experience in Machine Learning, Data, Software Engineering, and People. If he’s not working on a project, you’ll find him learning about new gadgets, writing code, taking photos, or running. Check out our ⁠show notes⁠ for more info on Duarte Carmo. --- Time Stamps (00:00:00) Trailer (00:01:07) Who is Duarte? (00:02:57) Use case: building a multimodal search engine (00:08:23) Challenges of developing and using a multimodal search engine (00:12:23) Human + machine: precision, reasoning and language (00:16:34) Challenges of communicating between humans and machines: freedom vs restriction, bias vs. variance (00:19:13) Why is it important to ask your client what their problem is? (00:22:58) Building an AI product: iterative approach vs. perfect launch (00:26:49) Priorities in AI development (00:31:29) Are customized AI solutions affordable or too expensive for small and medium businesses? (00:36:42) Privacy as a concern for multimodal search engines; APIs and privacy (00:38:56) Duarte’s achievement of building a customized multimodal search engine (00:40:14) Final remarks and book recommendations --- More on G.M.S.C. Consulting Follow us on our socials: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠LinkedIn⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠YouTube⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Book an appointment⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ with us. ⁠⁠⁠⁠⁠⁠Sign up to our newsletter⁠⁠⁠⁠⁠⁠. --- Music credits: storyblocks.com Logo credits: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Joshua Coleman, Unsplash⁠⁠⁠

If you asked AI to solve a Rubik’s cube, wouldn’t it be nice to understand all the steps it took to achieve the goal? In this episode of the AI and Digital Transformation Podcast by G.M.S.C. Consulting, you will meet the other half of  AI: symbolic AI. It gives a step-by-step procedure that leads to its final answer. As pointed out by Abzu’s founder and CEO, Casper Wilstrup, symbolic AI helps pharmaceutical, financial, manufacturing, and logistical companies solve hard problems with confidence.  Listen to this episode if you’re interested on answering the WHY to your unsolvable business problems using trustworthy and explainable AI. Who is Casper Wilstrup? Casper is the founder and CEO of Abzu®, the Danish/Spanish research startup that builds trustworthy AI to tackle high-risk challenges. Casper is the inventor of the QLattice® symbolic AI algorithm, an explainable AI that rationally reasons and makes evidence-based decisions. He has 20+ years of experience building large scale systems for data processing, analysis, and AI, and is passionate about the impact of AI on knowledge work and the intersection of AI with philosophy and ethics. Check out our show notes for more info on Casper and Abzu. Time Stamps (00:00:00) Trailer (00:01:22) About Casper and Abzu (00:03:44) The history behind AI, the challenge of doing symbolic AI (00:06:49) Why was Abzu formed? (00:08:42) Duality of AI: symbolic and sub-symbolic AI (00:13:05) How does symbolic reasoning work? (00:14:44) How did Abzu solve the problem of symbolic AI? Meet QLattice (00:18:43) Hypothetical scenario: Assessing legal cases (LLM vs. symbolic AI) (00:23:13) Minimum requirements to properly run symbolic AI  (00:24:42) Does symbolic AI need a lot of data to give a sound judgement? (00:27:37) Is symbolic AI similar to causal inference?  (00:33:53) Are there any limits to using symbolic AI? Explainable AI vs. Blackbox modelin AI (00:38:30) Should people have prior knowledge on symbolic AI to use QLattice? (00:41:15) Challenges and potentials of using symbolic AI (00:43:41) Abzu’s business model; Is QLattice open-source? (00:47:55) QLattice use case: medical research, life science and pharma (00:51:50) Logistics use case: does QLattice work on-demand? (00:54:15) Closing remarks & book recommendation --- More on G.M.S.C. Consulting Follow us on our socials: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠LinkedIn⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠YouTube⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Book an appointment⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ with us. ⁠⁠⁠⁠⁠Sign up to our newsletter⁠⁠⁠⁠⁠. --- Music credits: storyblocks.com Logo credits: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Joshua Coleman, Unsplash⁠⁠⁠

How many times did you lose customers because you address them with your marketing campaign? You sent so many freebies and promotions that they actually got annoyed and leave. And how many times did you actually waste money in freebies and promos on customers that would have bought from you anyway? The answer usually is a lot, but there is an answer to this problem, and this answer is called uplift modeling,an innovative machine learning technique that relies heavily on causal learning and causal inference, an innovative AI concept that leverages causal relationships between causes and effects, and not just simple correlations like 99% of machine learning out there. In this episode of the AI and Digital Transformation Podcast by G.M.S.C. Consulting, we are going to discuss the ins and out of this novel and powerful approach with Aleksander Molak, author of the book, Causal Inference and Discovery in Python. Stay tuned, because if you are a marketer, this is a game changing technology. Who is Aleksander Molak? Aleksander Molak is a Machine Learning researcher, educator, consultant, and author who gained experience working with Fortune 100, Fortune 500, and Inc. 5000 companies across Europe, the USA, and Israel, designing and building large-scale machine learning systems. On a mission to democratize causality for businesses and machine learning practitioners, Aleksander is a prolific writer, creator, international speaker and the author of a best-selling book Causal Inference and Discovery in Python. He’s a founder of Lesprie.io – a company that provides machine learning trainings for corporate teams, and the leader of CausalPython.io community. Aleksander has provided workshops, talks, and trainings for companies across industries, including market leaders like Mercedes-Benz, innovative disruptors like TechHub, international consulting companies like Lingaro and more. Check out our show notes to know more about Aleksander, his book, and his work. Timestamp (00:01:29) Who is Aleksander? (00:05:06) Machine learning vs. causal learning (00:12:27) Potential application of using causal learning - uplift modeling (00:22:45) Executing a causal learning project requires meaningful data (00:26:14) Causal learning as a low hanging fruit for businesses (00:29:41) Evaluating causal models (00:35:19) Balancing the cost of experimentation (00:38:26) Other applications of uplift modeling (00:41:13) Preparing for a causal learning project with Lespire Consulting (00:44:03) Educating data science teams on causal learning (00:46:53) When should someone buy uplift modeling? (00:50:07) Concluding remarks and book recommendation --- Follow us on our socials: ⁠⁠⁠⁠⁠LinkedIn⁠⁠⁠⁠⁠ ⁠⁠⁠⁠⁠YouTube⁠⁠⁠⁠⁠ ⁠⁠⁠⁠⁠Book an appointment⁠⁠⁠⁠⁠ with us. Sign up to our newsletter. --- Music credits: storyblocks.com Logo credits: ⁠⁠⁠⁠⁠Joshua Coleman, Unsplash⁠