rocher

Hi,I'mDurandeKamga,
Data scientist

Based in France

I build efficient backend systems and develop AI models with a focus on interpretability.
Build fast, design to convert, ready to grow!

Projects that I have worked on, showcasing my skills in web development and data science.

ChatDoctor screenshot 1
ChatDoctor screenshot 2
ChatDoctor screenshot 3
ChatDoctor screenshot 4

01 - ML / AI

ChatDoctor

An end-to-end medical assistant built on LLaMA 3.2-3B fine-tuned with QLoRA on 112k real doctor-patient consultations. Features full XAI explainability via SHAP and Integrated Gradients, streaming inference, FR/EN bilingual support, and live evaluation metrics (ROUGE-L, BERTScore F1), giving clinicians transparent, interpretable responses they can audit.

FastAPINext.jsLLaMA 3.2QLoRAPEFTSHAPCaptumTailwindCSSDocker
RAG, Finances publiques screenshot 1
RAG, Finances publiques screenshot 2
RAG, Finances publiques screenshot 3

02 - ML / AI

RAG, Finances publiques

A production-grade RAG system over 94 official French fiscal documents coolected using web scraping(Assemblée Nationale, Sénat, impots.gouv.fr, service-public.fr). Combines dense retrieval via pgvector with BM25 sparse search fused through Reciprocal Rank Fusion, cross-encoder reranking (mxbai-rerank-large-v1), and streaming generation using llama:8b. Features an inline PDF viewer with page-level citation linking, multi-source attribution, and full observability through Langfuse tracing.

FastAPINext.jsPostgreSQLpgvectorBM25RRFshadcn/uiCerebrasHuggingFace TransformersLangfuseTailwindCSS
Assurance ML screenshot 1
Assurance ML screenshot 2
Assurance ML screenshot 3
Assurance ML screenshot 4

03 - ML / AI

Assurance ML

An end-to-end machine learning platform to automate insurance fraud detection. It use XGBoost/Decision tree/Random forest/MLP with full SHAP explainability, counterfactual generation via DiCE, Anchors and LORE, enabling insurers to understand, audit, and challenge every model prediction with transparent, interpretable AI.

FastAPINext.jsXGBoostscikit-learnSHAPTailwindCSSDocker
DASMA Drift screenshot 1

04 - ML / AI

DASMA Drift

A research tool developed at LIMOS Lab for concept drift detection in time-series anomaly detection pipelines. It benchmarks adaptive drift strategies using deep learning models built with TensorFlow and the Aeon library, contributing to robust ML systems in production.

FlaskTensorFlowAeon
Farmadi API screenshot 1
Farmadi API screenshot 2

05 - Dev

Farmadi API

A production-grade REST API powering the Farmadi agricultural platform. Built with Django REST Framework, it handles asynchronous task processing with Celery and Redis, containerized with Docker, and backed by PostgreSQL, designed for scalability and reliability in agri-tech operations.

DjangoDRFDockerPostgreSQLRedisCelery
Gantt Builder screenshot 1
Gantt Builder screenshot 2

06 - Dev

Gantt Builder

An interactive browser-based Gantt chart builder for visualizing project timelines without any backend. Drag-and-drop task management, milestone tracking, and live export, built entirely with Chart.js and Tailwind CSS for a frictionless planning experience.

HTMLTailwindChart.js
Infolab Tech screenshot 1
Infolab Tech screenshot 2

07 - Dev

Infolab Tech

The official website for Infolab-Tech, a technology innovation company. Designed with a clean, professional aesthetic, it presents the company's services, expertise, and vision, built to convert visitors into leads through clear messaging and an intuitive layout.

HTMLBootstrap
Infox Detector screenshot 1
Infox Detector screenshot 2

08 - ML / AI

Infox Detector

A real-time fake news detection web app powered by classical NLP. It preprocesses articles with NLTK and runs them through a trained scikit-learn classifier, surfacing credibility predictions directly in the browser via a lightweight Django interface.

DjangoNLTKScikit-learn

I handle everything from design to development – without endless meetings. Saving you time and headaches.

Profile

I'm a web developer and data scientist ready to contribute to dynamic projects. My goal is to build functional and data-driven web solutions that prioritize tangible outcomes for users and businesses

From concept to launch, I handle every step of the process, ensuring it's smooth and hassle-free, so you can focus on your business.

Beyond tech, I'm into anime, running, and gaming.
Let's create something awesome!

Profilefuzz-ieee
Web & Mobile
  • Django logoDjango
  • FastAPI logoFastAPI
  • React logoReact
  • Svelte logoSvelte
  • React Native logoReact Native
AI & ML
  • Deep Learning logoDeep Learning
  • Machine Learning logoMachine Learning
  • Explainable AI logoExplainable AI
  • Time series analysis logoTime series analysis
Tools
  • Docker logoDocker
  • PostgreSQL logoPostgreSQL
  • Git/Github logoGit/Github
  • Celery logoCelery
  • Trello logoTrello

My career journey and technical expertise

  • IMT Nord Europe & Clermont Auvergne University logo
    IMT Nord Europe & Clermont Auvergne University logo

    November 2023 - Present

    Machine Learning Junior Researcher

    XAI for time series classification

    IMT Nord Europe & Clermont Auvergne University

  • LIMOS logo

    March 2023 - September 2023

    Machine Learning Junior Researcher

    Multivariate datastream and anomaly detection by trying to detect concept drift

    LIMOS

  • InfoLab Technology logo

    August 2021 - 2023

    Backend Developer

    InfoLab Technology

  • Gigalitic logo

    March 2022 - 2023

    Backend and Machine Learning Developer (Freelance)

    Gigalitic

  • SEED SARL logo

    April 2021 - October 2021

    Assistant Trainer

    SEED SARL