Dr Riad Ibadulla

About

Dr Riad Ibadulla

I am a Lecturer in Computer Science at City St George’s, University of London, where I teach deep learning, computer vision and programming, and research medical image analysis.

My PhD, at City, University of London, was on deep learning architectures for free-space optical AI accelerators. It produced FatNet, a method for adapting conventional networks to optical hardware by trading channel depth for spatial resolution, and led to Fat-U-Net for segmentation and ConvShareViT, a Vision Transformer that runs using only convolutions. After the PhD I worked on neural networks for intrusion detection that combine deep learning with formal methods, which led to PROTECTION.

Before the PhD I was a Deep Learning Engineer at Optalysys, which develops optical co-processors for AI. I designed and benchmarked models for high-throughput inference against GPU baselines, and contributed to a demonstration project with the UK National Cyber Security Centre on video-based extremism detection.

Download CV (PDF) Contact
Riad Ibadulla in a dark suit and blue tie, arms folded, standing beside a white column.

Teaching

At City St George’s, University of London, since January 2025:

  • Deep Learning for Image Analysis (MSc), module leader
  • Python Programming (Foundation year), module leader
  • Computer Vision (MSc)
  • C++ (second year)
  • Java (first year)
  • Supervision of MSc projects in deep learning and BSc final-year dissertations
  • Assessment design, lectures and lab support for other machine learning and software engineering modules

From 2021 to 2024 I was a Teaching Assistant on undergraduate and postgraduate modules in Python, C++, Big Data (GCP and PySpark) and Computer Vision.

Experience

  1. 2025 to present

    Lecturer in Computer Science

    City St George’s, University of London

    Teaching and supervision as above. Research on medical imaging: segmentation, alignment and landmark location in knee images.

  2. 2024 to 2025

    Postdoctoral Researcher (part-time)

    City, University of London

    Robust neural networks for intrusion detection, combining deep learning with SMT-based formal methods. Built evaluation tooling for controlled experiments, ablations and robustness under perturbation.

  3. 2021 to 2024

    Teaching Assistant

    City, University of London
  4. 2019 to 2020

    Deep Learning Engineer

    Optalysys

    Models for optical AI acceleration, benchmarked against GPU baselines under strict compute, memory and latency limits.

  5. 2019

    Python Developer (internship)

    Formation MSK Ltd

    Data analysis and decision-tree logic for a KPI analytics tool used with telecom retail clients.

Education

  1. 2020 to 2024

    PhD in Computer Science

    City, University of London

    Thesis: High Resolution Capabilities of Free-space Optical Neural Networks.

  2. 2018 to 2019

    MSc in Artificial Intelligence, Merit

    University of St Andrews

    Thesis: Simplex Optimisation for Aerial Image Stitching (code).

  3. 2015 to 2018

    BEng in Computer Systems Engineering, First Class

    City, University of London

    Thesis: Estimating Hyperparameters of Long Short-Term Memory in Speech Recognition Systems.

Recognition

First place, Generative AI Hackathon, 10 Downing Street Data Science team (July 2023)

With Team Redux, a team of civil servants and myself, I built a prototype that used a large language model to modernise the ministerial red box. We presented it to Minister Alex Burghart at 10 Downing Street, and I was invited to the meeting that set up the follow-on project. The government’s Incubator for AI went on to develop the concept into Redbox, which was used by more than 6,000 civil servants. City St George’s news coverage.

State scholarship, “Education of Azerbaijani Youth Abroad” (2014)

A fully funded government scholarship for study abroad, awarded on academic merit.

Open-source code

  • simulator

    A PyTorch simulator of free-space optical convolutional neural networks, using the angular spectrum method to model light propagation through a 4f system.

  • FatSpitter

    Takes a PyTorch sequential model and converts it into a FatNet.

  • FatUnet

    Fat-U-Net segmentation of the Oxford-IIIT Pet dataset and HeLa cell nuclei.

  • PROTECTION

    Code for the provably robust IoT intrusion detection system.

  • SmartRedBox

    The prototype built during the 2023 Generative AI Hackathon.

  • Anonymise_LLM

    A browser extension that uses local language models, via Ollama, to find sensitive information in text typed into chat assistants and replace it with generated substitutes.

  • DDK

    A GPT-style decoder-only transformer built from scratch in PyTorch: attention, positional embeddings, training loop, checkpointing and text generation.

Languages

English (fluent), Azerbaijani (native), Russian (native), Turkish (fluent).

University email for academic enquiries: [email protected]