Independent Researcher · Dakar, Senegal
AI Researcher in Mathematical Reasoning & LLM Evaluation
I study how language and multimodal models reason, where current evaluations break down, and how rigorous, verifiable benchmarks can reveal their limitations.
My work combines olympiad mathematics, proof auditing, benchmark design, scientific evaluation, and software engineering.
Can a system tell an invalid step from one that is merely compressed?
Studying whether reasoning systems can produce, inspect, and repair mathematically valid arguments rather than merely plausible final answers. The interesting cases are not arithmetic slips but arguments that read well and quietly depend on something never established.
What has to be true of a task before a score on it means anything?
Designing original, hand-verified tasks and evaluation procedures that expose model limitations while controlling for ambiguity, contamination, and weak answer checking. Difficulty and validity are separate properties, and a hard benchmark can still measure the wrong thing.
When a task requires reading a figure and reasoning quantitatively, which part actually fails?
Evaluating how models integrate textual, visual, and quantitative evidence in scientific problem-solving settings, including tasks constructed so that neither the image nor the text alone is sufficient.
What does rigorous evaluation look like for languages that mainstream benchmarks barely cover?
Building and evaluating systems for African languages and contexts that are often poorly represented in mainstream AI benchmarks, where the gap is both in model quality and in the measurement apparatus itself.
Research question
Can language-model-based mathematical critics reliably detect, localize, explain, and repair unsupported load-bearing claims while distinguishing them from ordinary mathematical errors and acceptable omitted details?
PremiseGuard investigates a failure mode in which a proof silently relies on a nontrivial claim that has not been established under the stated assumptions. The project studies controlled proof variants, earliest-error localization, missing-premise reconstruction, dependency analysis, and proof repair.
Detecting Unsupported Load-Bearing Claims in Mathematical Proofs
Named entity recognition for 21 African languages
An interactive mathematics and science learning environment
A French-language digital literacy platform
5 min read
In this blog post, I share my journey in building a FastAPI auth starter code, highlighting the common dilemma software engineers face: whether to use an existing library or build a custom solution. While libraries like fast-sso offer quick and reliable solutions, they may not always fit the unique needs of every project. Recognizing this, I created a FastAPI auth starter that provides the best of both worlds—quick setup and flexibility. This starter code is designed to help developers save time and adapt to various types of applications, from small projects to complex SaaS products. Whether you need basic authentication or more advanced roles and permissions, this starter code can give you a head start. The post emphasizes the importance of choosing the right tool for your project, team, and goals, and encourages developers to consider both libraries and custom solutions based on their specific needs.
4 min read
This blog shares my experience of learning from YouTube tutorials. I discovered that instead of following every step, focusing on the tools, understanding best practices, and solving challenges helped me grow faster as a developer. It’s not about copying tutorials, but applying the knowledge to real projects.
7 min read
This blog dives deep into the Machine Learning Life Cycle, outlining the essential stages from problem definition to model deployment and monitoring. It emphasizes the importance of clear problem identification, high-quality data collection, and robust model training. Each stage is critical to building scalable and reliable ML systems, with insights into selecting the right algorithms, avoiding common pitfalls like overfitting, and ensuring long-term model performance through continuous monitoring.
Alimoudine Idrissou is an independent AI researcher and software engineer based in Dakar, Senegal. His work focuses on mathematical reasoning, language-model evaluation, scientific multimodal benchmarks, and reliable evaluation methodology. He has designed and hand-verified original olympiad-level mathematics problems and scientific evaluation tasks for frontier AI systems. His background combines mathematics, computer science, software engineering, and mathematics education. He holds an M.S. in Computer Science from ESMT Dakar and a B.S. in Mathematics and Computer Science from Université d'Abomey-Calavi.
I read messages about evaluation methodology, mathematical reasoning, and benchmark design.