Mohamed Leila, PhD
I am a machine learning engineer based in Montreal. For ten years I have built ML platforms, evaluation pipelines, and responsible AI systems at Axon, ServiceNow, Rio Tinto, and Grownomic, a startup of my own that I started and shut down inside a year.
Now I run Adaptive Gradients, a consulting practice helping teams take AI from prototype to production. The wider question I keep coming back to: how AI changes the work, and what that means for the people who do it.
What I Do
All areas →Responsible AI, Trust & Governance
Building AI systems that are fair, transparent, and accountable from day one.
Privacy-Preserving ML
Federated learning, differential privacy, and adversarial robustness for enterprise data.
AI Adoption & Transformation
Helping organizations go from AI ambition to measurable production impact.
Agentic AI & LLM Applications
Designing and building autonomous AI workflows with human-in-the-loop oversight.
Featured Projects
View all →PPC Landing Page Builder
2025A chat-based platform that generates production-ready landing pages from natural language prompts. Built at Grownomic AI as sole engineer.
LangGraph · Agentic AI · Next.js · Supabase · Fly.io · Aider
ASR Fairness Evaluation Pipeline
2023Built an end-to-end pipeline to evaluate speech recognition providers across languages, accents, and demographics for law enforcement use cases. Published as FairLENS.
Speech Recognition · Fairness · Evaluation · Python
Federated Learning for Domain-Specific Language Models
2022Led an applied research project to train BERT on customer data without data commingling using federated learning. Validated technical feasibility and ran adversarial privacy evaluations.
Federated Learning · NLP · BERT · Privacy · Python
Where I Have Worked
Full background →
Recent Writing
View all →The Reliability Gap in AI Essay Grading
Why LLM-based essay graders score the same essay differently each run, what the MCAS rescoring incident reveals about the category, and the five engineering controls that turn a language model into a reliable scoring instrument.
Grownomic: Starting and Shutting Down an AI Startup in a Year
Three pivots, paying customers, and a platform policy that ended the final bet. What I learned building and shutting down Grownomic AI.
Want to work together?
I am available for AI consulting engagements. Let me know what you are working on.