Open to opportunities

zakaria kounima

@zakariakounima

Message

Data scientist and AI engineer specializing in LLM benchmarking, RAG pipelines, and NLP.

Morocco

Message

What I'm looking for

I seek hands-on roles building and optimizing LLMs and NLP systems, with opportunities to impact inference efficiency, deployment, and evaluation in collaborative, product-focused teams.

I am a data scientist and AI engineer focused on LLM benchmarking, NLP, and production-ready ML systems. I research hardware and software trade-offs for efficient inference and build pipelines that move models from experiment to deployment.

At Smartly.ai I led LLM GPU benchmarking across multiple GPUs with vLLM, developed a Streamlit benchmarking app, and analyzed model performance under concurrent user loads to guide infrastructure planning. I monitor hardware usage per model to inform production decisions.

I design and implement end-to-end NLP systems, including local-language models, fine-tuning workflows, and integrations of TTS/STT using Coqui-TTS and NeMo. I built a modular RAG pipeline with Elasticsearch retrieval, chunking strategies, and an evaluation framework comparing embedding models by cosine similarity, latency, and BLEU/ROUGE.

I also deliver practical ML products: a FastAPI backend for translation and voice-enabled web tools, a CNN-based plant-disease classifier integrated into an API and UI, and a Power BI sales insights dashboard. I embrace reproducible tooling, cloud deployment, and iterative evaluation to drive impact.

Experience

Work history, roles, and key accomplishments

Current

Data Scientist

Current

Smartly.ai

Dec 2024 - Present (9 months)

Conducted LLM GPU benchmarking and optimization across multiple GPUs using vLLM, analyzing model performance under concurrent user loads to guide infrastructure planning and reduce inference costs. Designed a Streamlit benchmarking app and integrated fine-tuned translators, TTS, and STT into a FastAPI cloud backend to enable local language NLP workflows.

LLM Benchmarking VLLM fastAPI Streamlit Transformers Coqui TTS NeMo

Data Scientist Intern

National School of Mines of Rabat

Jun 2022 - Aug 2022 (2 months)

Developed a CNN-based deep learning model to classify rose leaf diseases into three categories, applied data augmentation to improve generalization, and deployed the model via a FastAPI backend with a user-facing web interface.

CNN TensorFlow fastAPI Python Jupyter Deployment

Education

Degrees, certifications, and relevant coursework

National School of Mines of Rabat

Master of Science, Data Science

2019 - 2023

Activities and societies: Developed projects including CNN-based image classification for plant disease detection and backend integration using FastAPI; research and practical work in NLP and LLM benchmarking.

M.Sc. in Data Science from the National School of Mines of Rabat with coursework and projects in machine learning, deep learning, and applied data science from 2019 to 2023.

Mohammedia High School (CPGE)

CPGE, Preparatory Scientific Studies

2016 - 2019

Activities and societies: Completed coursework preparing for engineering school entrance; foundational training in mathematics, physics, and computer science.

Preparatory classes (CPGE) at Mohammedia High School focusing on foundational scientific and technical coursework from 2016 to 2019.