Mohamed Atoui
BSc Computer Science & AI
ML Engineer Co-Founder at Nordlys Labs
About
Hi, I'm Mohamed, Co-Founder & ML Research Lead at 
Nordlys Labs, building intelligent LLM routing systems that select the best model for each task — reducing inference costs by 50–60% and scaling to 500+ users.
I'm a 3rd-year Computer Science & AI student at 
Royal Holloway, University of London with First-Class Honours. My work spans ML research (transformer attention mechanisms, adversarial robustness), data science (sports analytics at 
Stats Perform), and teaching.
You can find more about my background in my CV.
Interests: Machine Learning, LLM Systems, Applied AI, Product Development.
Latest News
Presented MoM Framework at Online Meetup
Presented MoM Framework at Online Meetup
Presented the MoM (Mixture of Models) framework in an online meetup, followed by engaging discussions with the audience on LLM routing and model selection strategies.
Teaching Assistant at Royal Holloway
Teaching Assistant at Royal Holloway
Started as a Teaching Assistant for Object-Oriented Programming & Software Engineering at Royal Holloway, University of London.
Product Data Intern at Stats Perform
Product Data Intern at Stats Perform
Joined Stats Perform as a Product Data Intern, working on sports analytics and data science.
ML Research at ISG Smart Card & IoT Security Center
ML Research at ISG Smart Card & IoT Security Center
Started ML research at the ISG Smart Card & IoT Security Center, working on adversarial robustness in ML-based intrusion detection for autonomous vehicles.
Founded Nordlys Labs
Founded Nordlys Labs
Co-founded Nordlys Labs, building intelligent LLM routing systems that select the best model for each task. Beat frontier models on SWE-bench benchmarks. Scaled to 500+ users.
Check out my latest work
Nordlys Labs
Nordlys Labs
Intelligent LLM routing system that selects the best model for each task. Outperformed frontier models on SWE-bench benchmarks, reduced inference costs by 50–60%, and scaled to 500+ users.
ZeroToDiffusion
ZeroToDiffusion
From-scratch implementation of diffusion models: derived the probability theory, built a VAE, then implemented DDPM step by step to understand image generation from first principles.
KV-cache optimization
KV-cache optimization
Empirical comparison of MHA, MQA, and GQA attention mechanisms, measuring perplexity vs KV-cache size trade-offs on TinyStories. GQA achieved 4x cache reduction with minimal quality loss.
Skills
ML & AI
Backend
Frontend
Infrastructure
Internship & Work Experience
Nordlys Labs
Stats Perform
ISG Smart Card & IoT Security Center
Royal Holloway, University of London
Awards & Honors
Scaled Nordlys Labs to 500+ users
Academic Services
Object-Oriented Programming & Software Engineering
Get in Touch
Want to chat? Feel free to reach out via email →
- • Ask questions
- • Explore collaboration opportunities