Welcome to our new “Game Changers in Tech” column, where we will highlight the brilliant minds at Bitrock and their commitment to tackling some of the most innovative technological challenges of our time. In a constantly evolving field like IT, staying ahead is not just a necessity, it is an art. Our Bitrockers embody this philosophy, mastering cutting-edge technologies and demonstrating excellent technical skills in adapting to new situations and addressing new technology trends.
In each interview, we will uncover the stories of those who make Bitrock a leader in innovation. We will explore their experiences, the challenges they have overcome, and the skills they have developed to meet the needs of our clients and deliver ground-breaking IT consulting projects. Through practical examples, we will understand how these exceptional individuals are preparing to meet the demands of the marketplace and lead our company into the future.
Our first “game changer” is Giovanni Vacanti, Data Scientist, who passed through Singapore and many other places in the world and landed at Bitrock to enrich our Data, AI & ML Engineering team.
Let’s get to know him better!
Can you explain your position and career path?
After obtaining a Master’s Degree in Physics from the University of Palermo in 2007, my passion for understanding complex systems led me to pursue a Ph.D. in Physics at the Center for Quantum Technologies at the National University of Singapore, where I graduated in July 2013. I then continued my research in quantum physics as a postdoctoral fellow at the Institut Neel in Grenoble, France, until June 2015.
Following my time in academia and my interest in AI, I joined a London-based start-up focused on Machine Learning solutions for business in 2016. During my time there, which lasted almost 7 years, I contributed to the growth of the company and the AI community in the UK through various projects, including the development of the open-source library Alibi, Alibi-detect and seldon-core.
In 2022 I started a new phase of my career back in Italy by joining Bitrock. At Bitrock I work on various projects involving “traditional” ML models, such as forecasting or recommendation systems, but my main focus is on Generative AI powered by LLMs solutions.
What has been the most challenging technology issue you’ve faced at Bitrock?
In Machine Learning and AI projects, you always have to use advanced cutting-edge technologies i. The open source community makes research tools available for general use very quickly, so we always have to look for the most advanced tools.
The most difficult challenge I faced was a project where, given the client’s strict requirements (basically everything on-premise, so no ChatGPT API), we ended up fine-tuning a 3 billion LLM on a 4-GPUs cluster. I learned there that training or fine-tuning an LLM is harder than you might think, even for “small” models with a few billion parameters. And also expensive, because GPU time costs money.
On the other hand, OpenAI, Anthropic and more and more companies offer a paid API service to access their LLMs or their big artificial brains. This is much cheaper, both in terms of cost and effort, and open source tools like LangGraph make it very easy to harness the full power of these huge models.
What do you think are the most important trends for the future of our industry?
First, the advancement of generative AI, especially through large language models (LLMs), is already redefining problem solving across sectors. The same goes for ethical AI and accountability, as stakeholders demand transparency. And we should not forget that these trends will drive significant innovation in the coming years.
Another important trend is the evolution of GPU resources. Cloud-based GPU services allow organizations to access high-performance computing resources at a reasonable cost. In addition, innovations in GPU architectures are enabling more efficient processing, which is critical for meeting the computational demands of large-scale AI applications.
However, as LLM grows, deployment requires more and more GPU resources, which can be significantly more expensive and it may be more convenient to use paid API services such as OpenAI or Anthropic.
How do you keep up with these trends and how does Bitrock support you in this continuous learning?
As I said before, open source now keeps up with research almost in real time, and progress is exponential. This means that any new tool, software or technique can become obsolete in a few months.
My learning is always driven by the curiosity to explore new trends and technologies. I keep up to date by reading scientific articles and specialized blogs, as academic discoveries are now being rapidly implemented into production-ready open-source tools.
Bitrock supports this process by providing access to platforms such as Coursera and DeepLearningAI, with always up-to-date courses. For those who want to specialize, it’s essential to select reliable sources and develop a discerning mind, and Bitrock encourages this mindset by providing us tools and opportunities for continuous growth.
What advice would you give to those who want to keep up to date with AI?
My advice is to adopt a “continuous learning” approach. The open-source community is very active in providing free courses, blogs, and tools. I believe that the breakthroughs in natural language processing and generative AI that we have seen in recent years are triggering a paradigm shift in the way both researchers and software engineers work.
To stay ahead in this rapidly evolving landscape, it’s essential to embrace collaboration and share knowledge with others in the field. Of course, It’s important to establish a routine of learning and updating, perhaps dedicating a few hours a week to personal development. In my opinion the best way to do this is to learn “in the field” by using the most recent tools your project requires. For example, tools like LangChain did not exist 1 or 2 years ago, but I learned how to use it in a recent project and I found it extremely useful.
Additionally, actively participation in online communities, forums, and discussion groups can provide valuable insights and networking opportunities.
Thanks to Giovanni Vacanti, Senior Data Scientist @ Bitrock, for this interview.
If you are looking forward to meeting other Bitrockers and learning how they have changed their game… stay tuned!
