Memseek
An open-source context engine for AI agents. It updates knowledge as evidence changes, keeps its sources and history, and assembles context for each task.
I co-founded Memseek and build the engine that maintains this context.
RESEARCH & CONSULTINGPARIS, FRANCE
ZAKARI AIT OUAZZOU
AI research.ML engineering.
I’m Zakari, a PhD researcher at Télécom Paris and co-founder of Memseek. I research model explainability and build and evaluate machine-learning systems.
YOUR PROJECT, STEP BY STEP
Find the right problem.
I help define the decision, the data, and what success means.
PhD · Télécom Paris / Institut Polytechnique de Paris
Co-founder · Memseek
01 / SELECTED WORK
Research &
open source.
MemBukkit
I designed the memory and retrieval architecture: preserve source evidence, select what matters, and limit how much the model needs to read.
MemBukkit82.0%
Full context56.4%
Reported answer accuracy · LongMemEval-S · 500 questions.
Same
GPT-4o-mini reader and official GPT-4o judge.
BDA 2025 · XAI RESEARCH
MINES
I develop methods that update a model’s explanations as the model changes, reducing the need to recompute them from scratch.
MINES at BDA 2025Publication details
With Talel Abdessalem and Albert Bifet.
- MINES: Model-agnostic INcremental, Entropy-based Solutions for Efficient Local Explanations in Streaming Machine Learning. BDA 2025, Toulouse.
- MINES: Efficient Incremental Local Feature Attribution for Evolving Data Streams. Transactions on Large-Scale Data- and Knowledge-Centered Systems. Accepted; in production.
ONGOING RESEARCH
M-BOX
I’m developing models whose predictions follow an inspectable graph of concepts, and testing whether that structure remains interpretable as the data changes.
02 / CONSULTING
What I can
help you build.
01
Model explainability
Investigate what drives your model’s predictions, evaluate its explanations, and diagnose changes in its behavior.
02
Agent memory & retrieval
Build and evaluate memory and retrieval systems that keep answers connected to their sources.
03
Applied ML & evaluation
Develop forecasting or ranking models, design experiments, and build evaluation pipelines to compare approaches.
03 / INDUSTRY EXPERIENCE
Where I’ve
done the work.
AUG 2024–PRESENT
Darma
Machine Learning Engineer
Led applied ML for a scheduling assistant: preference-aware ranking, temporal knowledge graphs, retrieval, and evaluation.
FEB 2023–FEB 2024
Voltalis
Data Scientist / R&D Engineer
Built LSTM forecasting and diagnostic models for energy demand and demand response, working with noisy operational time series.
FEB–AUG 2022
TotalEnergies
Data Scientist Intern · Digital Factory
Developed explainability and drift-analysis methods for time-series forecasts. This industrial work later fed into my PhD research.
SEP 2021–FEB 2022
DEEPXP
Data Science Consultant
Built NLP and data-science applications for Airbus projects.
ZAKARI AIT OUAZZOU · PARIS, FRANCE
04 / ABOUT
I’m Zakari.
I’m doing a PhD in AI at Télécom Paris / Institut Polytechnique de Paris, supervised by Talel Abdessalem and Albert Bifet. My focus is making model explanations more interpretable and efficient.
I spent January–March 2026 at the University of Waikato in New Zealand. I also teach AI and logic at Télécom Paris and mentor computer-science projects at ENSTA Paris.
- PhD in Artificial Intelligence
- Télécom Paris / IP Paris · 2024–2027 expected
- MSc in Data Science & AI
- Sorbonne Université · 2022 · Excellence Diploma
- BSc in Computer Science & Mathematics
- Sorbonne Université · 2020 · Highest honors
Outside work, I climb.
LinkedInCONTACT
Let’s work
together.
Tell me what you’re building, where you’re stuck, and what you need help with.
Email me zak@aozakari.com