Empowering data science teams to succeed
$gleb --about
Hey there, I'm Gleb, and I'm passionate about empowering data science teams to succeed by building reliable and scalable MLOps solutions — because effective MLOps isn't just about tools; it's about aligning people, processes, and platforms to solve real business problems.
Outside of work, I co-organise the Agentic AI Foundation (AAIF) Community London — Europe's largest AI engineering meetup — and love sharing what I've learned through blogging and public speaking.
Before entering the tech industry, I conducted scientific research at world-class physics labs in the US and Switzerland, working on large-scale, multidisciplinary projects. In my spare time, I enjoy road cycling and running.
$
I am actively engaged in the Agentic AI Foundation (AAIF) Community London, the largest AI engineering community in Europe, as a co-host and events organiser. We are putting regular in-person meetup-ups, for the community to learn together and discuss latest agentic AI trends over pizza.
Turn a repository into a slide deck that tells you when it stops matching the code. SlideOps is a pair of Agent Skills: one builds a single-file HTML deck from a repo, the other checks that deck later. Every reference in the deck carries the file it came from, the exact line range, and a hash of those source lines, so the check is standard-library Python: no model, no network, zero tokens, milliseconds.
The repo is its own plugin marketplace, so in Claude Code it is two lines: add the marketplace,
/plugin marketplace add glukicov/slideopsthen install the plugin from it:
/plugin install slideops@slideopsAny agent with the skills CLI gets it in one line:
npx skills add glukicov/slideops
It also runs on Codex, Copilot CLI and
OpenCode: all three read
~/.agents/skills, and
./install.sh in a repo puts the
skills where each of them looks.
That is the question — the platform is the answer. How does a correct MLOps strategy enable success for data science projects in a large organisation? What is an effective MLOps team?
Are you an AI engineer trying to bridge local development and production deployment of ML projects? Find out how pipelines enable scalable experimentation with Docker and uv on Google Cloud.
In this video, I share my professional journey at Virgin Media O2, where we are utilising the power of Google Cloud to enable us to make data-driven decisions that improve our customers' experiences. For this work, our team won the DevOps award in 2022 and 2023.
In this interview, I share my journey from particle physics to data science and MLOps, and how I discovered a passion for building data platforms and helping teams work with data at scale. I talk about my path into the field, my proudest moments as an ML engineering manager, and the importance of good fundamentals, community, and clean data.
In this podcast, I discuss why your business might benefit from MLOps, MLOps Community London meetups and the exciting work we do at Electric Twin on simulating human behaviour in real-time with AI.
With Ubuntu, CUDA and port-forwarding. I cover how to set up secure remote access over ssh, wake-on-LAN feature to switch on your GPU server remotely and mounting the remote file system on your laptop.
Startups move fast, and as it turns out, you can move fast on security & compliance too!
Live conversation with Erica Hughberg about scaling MLOps at Electric Twin and what inspired my journey into data & AI.
The g − 2 experiment at Fermilab, near Chicago, discovered a tantalising sign of New Physics (a new force of nature!). This was done by measuring a deviation between the experimental and theoretically predicted value of the muon magnetic anomaly. As part of my PhD, I collaborated on the experiment with 200 scientists and engineers. In 2026, the experiment was recognised with the prestigious Breakthrough Prize in Fundamental Physics.
To accurately measure this anomaly, a calibration of the tracking detectors is required. The main project of my PhD involved developing, testing and deploying the calibration software framework, which increased the yield of data by 3% and the data quality by 4%. The alignment procedure is a statistical problem, which involved the optimisation of the p-values of tracks via matrix inversion.
There is also an additional measurement that will be made using the tracking detectors: setting a new limit on the electric dipole moment (EDM) of the muon, producing a world-leading result. I have been developing algorithms to measure the muon EDM, by analysing large and complex datasets using regression and Fourier transform methods.
Presenting at the American Physical Society conference in Boston. My talk highlighted the significance of precise tracking systems in high-energy physics experiments and their role in probing fundamental particle properties.
An essential component of the experiment is the data acquisition (DAQ) system, which manages the data flow from the detector electronics. The experiment is acquiring raw data at a rate of 20 GB/s. This is accomplished by employing a parallel data processing architecture using 28 high-speed GPUs (NVIDIA Tesla K40), reducing the recorded data rate to 200 MB/s. To support the smooth operation of the experiment and ensure continuous data taking, a team of data acquisition (DAQ) experts are available for 24/7 support. I have actively participated as the DAQ on-call computing expert between 2017 and 2019.
Delivering an alumni talk at the Woodhouse College about my all-time favourite particle - the muon - and how it can be used for treasure-hunting, making borders more secure, and unlocking the mysteries of the Universe. I come back to the college every year to encourage students in pursuing a career in STEM.
Guiding a public tour of the Fermilab Muon g — 2 experiment. The experiment has measured an anomaly in a property of a fundamental particle, expanding our knowledge of the Universe and providing evidence of the existence of new particles.
I wrote an article about building a personal Bitcoin access point with Raspberry Pi. Read to learn about using an Electrum node for interacting with the decentralised blockchain network.
Explaining spectroscopy to primary school students at the Your Universe astronomy festival at UCL. I was involved with the festival between 2010 and 2015. Every year, I was able to deliver educational activities to 300 pupils. I also had a chance to meet Geraldine Cox during the festival.
As part of a charity fundraising, I ran my first full marathon (5:31 / 1 km elevation) and cycled 100 km (4:44 / 2 km elevation). This motivated me all the way (twice!) to the top of that mountain, and our team manged to fundraise £2.3K for Street Child.
Next year, our team was back at it! With a mini tour-of-Netherlands: 160 km day 1 (6:28 / 0.8 km elevation) and 175 km day 2 (6:37 / 1.1 km elevation). This time, we have raised £3.8K for Street Child.
↑↓move↵openescclose