Mahdi Bedja

Mahdi Bedja

AI Engineer · Paris

I work on applied AI: medical imaging, speech recognition, LLM-based tools and computer vision on mobile devices. I graduated from ESI Algiers and I'm now doing a Master 2 in Intelligent Systems at Sorbonne Paris Nord.

Looking for a final-year internship (PFE) in applied AI, from March 2027
mahdibedja@gmail.com LinkedIn GitHub

About

For ten months I was an AI engineer at BM TECH, a company that makes software for medical laboratories and hospitals in Algeria. I worked on chest X-ray analysis, French medical speech recognition, real-time dictation, and an assistant that answers questions over patient records using locally hosted LLMs.

Before that, my engineering thesis was a mobile app that detects crop pests offline. I prepared the dataset, compared eight detection models, and shipped the best one inside an Android app with its own backend.

I like taking a model all the way to something people actually use, and I'm looking to keep doing that during my final-year internship.

Experience

AI Engineer, BM TECH

Nov 2025 – Aug 2026
Algiers, Algeria · Software for medical laboratories and hospitals
  • Detection, segmentation and classification of 14 lung pathologies on chest X-rays (YOLO, CNNs), plus a separate bone fracture detection pipeline.
  • French medical speech recognition: built a synthetic speech dataset by cloning physicians' voices with Qwen3 from real radiology reports, then fine-tuned Whisper. 3% WER on a held-out synthetic test set.
  • Real-time medical dictation on WhisperLiveKit / faster-whisper, with fuzzy matching and voice commands. 8% WER in live streaming.
  • Agentic RAG system with LangChain and locally hosted Qwen models to answer questions over patient records and clinical documents.
  • Semantic matching of lab test results across hospitals using dense embeddings of the LOINC dataset.
  • Packaged the services with Docker, with structured logging and monitoring.

This work belongs to the company, so the code isn't public.

Engineering Intern, CM Consulting

Jul – Aug 2024
Algiers, Algeria
  • Python tool that extracts and structures data from PDF tenders (spaCy, pdfplumber), with a Tkinter interface and automatic report generation.

Projects

The pest detection app showing a detected leafhopper with its confidence and treatment
Engineering thesis · 2024–2025

Mobile pest detection

An Android app that identifies 13 crop pests from a photo, offline, and suggests treatments. I curated a subset of the IP102 dataset, trained and compared YOLOv7 to v11, SSD and Faster R-CNN, and chose YOLOv11s (93.8% mAP@50) for its accuracy-to-size ratio. It runs on the phone with TensorFlow Lite, and a Django backend syncs each user's history.

PyTorch · Ultralytics · TensorFlow Lite · Kotlin · Jetpack Compose · Django · PostgreSQL

Legal monitoring platform with an LLM chatbot

Team project. A chatbot that answers questions about legal texts, on a Django and React platform that scrapes new laws with Selenium, indexes them in Elasticsearch, and notifies subscribers.

LLMs · Django · React · Elasticsearch · Selenium

PostgreSQL database with automated deployment

Database for an event ticketing platform, with its logic in PL/pgSQL functions, Flyway migrations, tests, and a Jenkins pipeline that deploys to Render.

PostgreSQL · Flyway · Jenkins

Parking reservation app

Android app for browsing parking lots across Algerian communes, backed by a Django REST API that manages parking places, availability and reservations.

Kotlin · Jetpack Compose · Django REST Framework

Word certificate generator

Small desktop tool that reads a list of people from a Word table and generates one filled-in certificate per person from a template.

Python · python-docx · Tkinter

Skills and education

Generative AI
Agentic RAG, LangChain, LLM agents, local LLM inference (Qwen), embeddings
Speech and vision
Whisper fine-tuning and streaming, ASR/TTS, YOLO, SSD, Faster R-CNN, segmentation
ML frameworks
PyTorch, TensorFlow, Keras, Hugging Face, TFLite, spaCy, OpenCV
Backend and mobile
Python, Django, REST APIs, PostgreSQL, Elasticsearch, Kotlin, Jetpack Compose
Infrastructure
Docker, Jenkins, GitHub Actions, Git, logging and monitoring

Master 2, Intelligent Systems

Sorbonne Paris Nord University · 2026 – present

Engineering degree, Computer Science

ESI Algiers · 2020 – 2025

Languages: English (C1, IELTS 7.5), French (C1, TCF), Arabic (native).

Contact

I'm looking for a final-year internship in applied AI starting March 1, 2027. The best way to reach me is by email at mahdibedja@gmail.com, or on LinkedIn.