Firas Tlili - Detection Transformers Fine Tuning for Custom Object Detection

Q: Where is Firas based?

Firas is based in Gafsa, Tunisia .

Q: What languages does Firas speak?

Firas speaks the following languages: Arabic (Native), English (Advanced) .

Q: How many years of experience does Firas have?

Firas has at least 4 years of experience. During this time, Firas has worked in at least 6 different roles and for 2 different companies . The average length of individual experience is 1 year and 9 months . Note that Firas may not have shared all experience and actually has more experience.

Q: What roles would Firas be best suited for?

Based on recent experience, Firas would be well-suited for roles such as: Detection Transformers Fine Tuning for Custom Object Detection, Fine-Tuning Large Language Models (LLMs) Efficiently with Unsloth + LoRA, Multimodal AI Agent for Enhanced Content Understanding with LlamaIndex, NVIDIA NIM, and Milvus .

Q: What is Firas's latest experience?

Firas's most recent position is Detection Transformers Fine Tuning for Custom Object Detection .

Q: What companies has Firas worked for in recent years?

In recent years, Firas has worked for Omdena.com , Intelligent Machines Lab and University of Gabes .

Q: Which industries is Firas most experienced in?

Firas is most experienced in industries like Healthcare , Agriculture , and Education . Firas also has some experience in Sport and Information Technology .

Q: Which business areas is Firas most experienced in?

Firas is most experienced in business areas like Research and Development , Product Development , and Project Management . Firas also has some experience in Information Technology .

Q: What is Firas's education?

Firas holds a Master in Automatic Electrical Engineering from University of Gabes .

Q: Does Firas have any certificates?

Firas has 8 certificates . Among them, these include: Generative Adversarial Networks (GANs) Specialization , Google Cloud Associate Cloud Engineer , and Google Cloud Professional Cloud Architect .

Recommended expert

Gafsa, Tunisia

Check rate

Experience

Sep 2025 - Sep 2025

1 month

Detection Transformers Fine Tuning for Custom Object Detection

Developed an end-to-end DETR-based bone fracture detection system using PyTorch Lightning, Hugging Face Transformers, Supervision, and OpenCV, fine-tuned on ~1,200 COCO-formatted X-ray images to detect five fracture classes, achieving over 90% precision, fast inference (<50 ms/image), and side-by-side visual validation for clinical interpretability.
Tools: Python, PyTorch, PyTorch Lightning, Hugging Face Transformers, Supervision, OpenCV, Roboflow, COCO dataset, Matplotlib, Git, Google Colab, Jupyter Notebook.

Sep 2025 - Sep 2025

1 month

Fine-Tuning Large Language Models (LLMs) Efficiently with Unsloth + LoRA

Developed an efficient fine-tuning pipeline for large language models using Unsloth and LoRA, integrating 4-bit quantization, parameter-efficient training, and end-to-end workflows — from chat-style dataset preparation to model evaluation and multi-format export — enabling scalable, high-performance customization of multi-billion parameter models on a single GPU.
Tools: Python, Unsloth, LoRA, PEFT, Hugging Face Transformers, TRL (SFTTrainer), 4-bit quantization, mixed precision, PyTorch, Google Colab, Jupyter Notebook, Git.

May 2025 - May 2025

1 month

Multimodal AI Agent for Enhanced Content Understanding with LlamaIndex, NVIDIA NIM, and Milvus

Built a multimodal retrieval-augmented generation system using LlamaIndex, NVIDIA NIM microservices, and Milvus for real-time document and image comprehension, enabling users to interactively query PDFs, PowerPoint, and image data via a Streamlit chat interface, integrating LLMs and VLMs for advanced document and visual content analysis.
Tools: Python, Streamlit, LlamaIndex, NVIDIA NIM, Milvus, DePlot.

Jan 2025 - Feb 2025

2 months

End-to-End Medical Chatbot with LLMs, LangChain, Pinecone, and LLMOps

Built an end-to-end production-ready medical chatbot using large language models, LangChain, and Pinecone with a retrieval-augmented generation pipeline, delivering accurate, privacy-compliant health responses and real-time medical data retrieval with integrated validation for clinical relevance and safety.
Tools: Python, Flask, OpenAI GPT, LangChain, Pinecone, GitHub Actions, Docker, AWS EC2, ECR.

Sep 2022 - Present

3 years 8 months

Machine Learning Engineer

Omdena.com

Led an Omdena Local Chapter in Tunisia, organizing and managing a team of volunteers to collaborate on social impact projects and mentoring individuals to develop skills in data science and machine learning.
Developed deep learning models using Python and TensorFlow to detect olive leaf disease with 95% accuracy, helping local farmers cut expenses by 20% through early intervention strategies.
Developed computer vision models for red blood cell classification to diagnose sickle cell disease that increased predictive accuracy by 25%, utilizing Python, TensorFlow, and OpenCV, and collaborating closely with data scientists and engineers from Benin.

Jan 2022 - Dec 2022

1 year

Gabes, Tunisia

Deep Learning Research Engineer

Intelligent Machines Lab, University of Gabes

Automated soccer highlight extraction using deep learning and a custom YOLOv7 model fine-tuned with PyTorch, achieving a 90% increase in detection accuracy for key events and reducing video processing time by 90%.
Developed an image extraction pipeline (Python), a video compression tool (OpenCV, FFMPEG), and a full-stack web app (Django, React) to condense and deliver 90-minute matches into 5-minute summaries, cutting editing time by 80% and enabling real-time access to highlights.
Drove cross-functional collaboration in a hybrid academic-industry setting, aligning research goals with technical execution to accelerate innovation in sports video analysis.

AI Football Video Analysis System with YOLO, OpenCV, and Python

Developed a full-scale, AI-powered football analysis system leveraging state-of-the-art computer vision, deep learning, and tracking algorithms to monitor player performance, calculate ball possession, and deliver real-time match insights, blending YOLOv11 (Ultralytics) for precise object detection with KMeans clustering, optical flow, and perspective transformation to ensure accurate, real-world analytics.
Tools: Python, OpenCV, NumPy, Pandas, Matplotlib, Scikit-learn, YOLOv11 (Ultralytics), optical flow, perspective transformation, Jupyter Notebook, VS Code, Git.

End-to-End Kidney Disease Classification with MLflow, DVC, and Cloud Deployment

Designed and deployed a full ML pipeline for kidney disease classification using MLflow for experiment tracking, DVC for data versioning and orchestration, and AWS EC2/ECR with GitHub Actions for CI/CD, achieving reproducibility, scalability, and robust cloud deployment of the ML model via Docker.
Tools: Python, Pandas, Scikit-learn, MLflow, DVC, Docker, GitHub Actions, AWS (EC2, ECR, IAM), Flask.

Real Time Human Activity Recognition Video Data Annotation Tool

Architected and delivered a professional PyQt5 desktop application integrating YOLOv11 pose estimation models with OpenCV for real-time multi-person annotation, featuring a scalable threaded video processing pipeline, comprehensive COCO-compliant data management with automated backups, multi-format export capabilities (YOLO, Pascal VOC, CSV), and a robust project workflow that significantly reduced training dataset creation time for human activity recognition detection in security and surveillance applications.
Tools: Python 3, PyQt5, YOLOv11, OpenCV, threading, Ultralytics, Roboflow, YOLOv11-Pose, NumPy, JSON, XML, modular design, real-time video annotation.

Industry Experience

See where this freelancer has spent most of their professional time.

Experienced in Healthcare, Agriculture, Education, and Sport.

Healthcare

Agriculture

Education

Sport

Business Area Experience

See which departments and functions this freelancer has contributed to most.

Experienced in Research and Development, Product Development, Project Management, and Information Technology.

Research and Development

Product Development

Project Management

Information Technology

Summary

AI and Machine Learning expert with over 7 years of experience in developing and deploying cutting-edge deep learning models, AI pipelines, and full-stack ML solutions across domains including medical imaging, agriculture, and multimedia. Skilled in Computer Vision, Natural Language Processing, and Generative AI, with hands-on expertise in Python, PyTorch, TensorFlow, Keras, OpenCV, and cloud platforms (AWS, Google Cloud, Azure). Proficient in data preprocessing, feature engineering, model evaluation, hyperparameter tuning, and deploying production- ready AI systems using Docker, Kubernetes, MLflow, and DVC. Known for building optimized, high-performance AI models, fine-tuning large language models, and delivering scalable, innovative solutions that drive measurable impact and solve complex real-world problems.

Languages

Arabic

Native

English

Advanced

Education

Oct 2020 - Jun 2022

University of Gabes

Master of Automatic Electrical Engineering · Automatic Electrical Engineering · Gabes, Tunisia

Certifications & licenses

Generative Adversarial Networks (GANs) Specialization

Google Cloud Associate Cloud Engineer

Google Cloud Professional Cloud Architect

Google Cloud Professional Data Engineer

Google Cloud Professional Machine Learning Engineer

Huawei Certified ICT Associate: Artificial Intelligence

Machine Learning Specialization by Stanford University

Microsoft Certified: Azure Data Scientist Associate

Statistics

Experience

Total positions 9

Experience in Healthcare 3.5 y

Avg length 6 m

Longest experience 3 y 7 m

Global Experience

Countries worked in 1 (Tunisia)

Primary country Tunisia

Expertise

Recent roles Detection Transformers Fine Tuning for Custom Object Detection, Fine-Tuning Large Language Models (LLMs) Efficiently with Unsloth + LoRA, Multimodal AI Agent for Enhanced Content Understanding with LlamaIndex, NVIDIA NIM, and Milvus

Main industries Healthcare, Agriculture, Education

Main business areas Research and Development, Product Development, Project Management

Qualifications

Highest degree Master

Certifications earned 8

Profile

Created

October 2025

Last Update

October 2025

Need a freelancer? Find your match in seconds.

Try FRATCH GPT

Frequently asked questions

Do you have questions? Here you can find further information.

Where is Firas based?

Firas is based in Gafsa, Tunisia.

What languages does Firas speak?

Firas speaks the following languages: Arabic (Native), English (Advanced).

How many years of experience does Firas have?

Firas has at least 4 years of experience. During this time, Firas has worked in at least 6 different roles and for 2 different companies. The average length of individual experience is 1 year and 9 months. Note that Firas may not have shared all experience and actually has more experience.

What roles would Firas be best suited for?

Based on recent experience, Firas would be well-suited for roles such as: Detection Transformers Fine Tuning for Custom Object Detection, Fine-Tuning Large Language Models (LLMs) Efficiently with Unsloth + LoRA, Multimodal AI Agent for Enhanced Content Understanding with LlamaIndex, NVIDIA NIM, and Milvus.

What is Firas's latest experience?

Firas's most recent position is Detection Transformers Fine Tuning for Custom Object Detection.

What companies has Firas worked for in recent years?

In recent years, Firas has worked for Omdena.com, Intelligent Machines Lab and University of Gabes.

Which industries is Firas most experienced in?

Firas is most experienced in industries like Healthcare, Agriculture, and Education. Firas also has some experience in Sport and Information Technology.

Which business areas is Firas most experienced in?

Firas is most experienced in business areas like Research and Development, Product Development, and Project Management. Firas also has some experience in Information Technology.

What is Firas's education?

Firas holds a Master in Automatic Electrical Engineering from University of Gabes.

Does Firas have any certificates?

Firas has 8 certificates. Among them, these include: Generative Adversarial Networks (GANs) Specialization, Google Cloud Associate Cloud Engineer, and Google Cloud Professional Cloud Architect.

What is the availability of Firas?

Firas is immediately available for suitable projects.

What is the rate of Firas?

Firas's rate depends on the specific project requirements. Please use the Meet button on the profile to schedule a meeting and discuss the details.