Refik Malli
Vertical Technical Leader at OROBIX
Milan, Lombardy, Italy
Summary
🤩
Rockstar🎓
Top SchoolRefik Malli is a Vertical Technical Leader with 11 years of experience building and shipping production-grade deep learning and agentic AI systems, currently leading AI product development at Orobix in Milan. He specializes in combining vision-language models, LLMs, OCR, and layout/table extraction to unlock unstructured information from complex technical documents and power intelligent search and retrieval. Previously he delivered CV solutions across pharmaceutical, manufacturing, and agriculture domains—ranging from anomaly detection and segmentation to reinforcement learning agents—while also creating reproducible ML tooling like the open-source Quadra experiment framework. Hands-on in MLOps and deployment (PyTorch, FastAPI, Docker, Azure), he architects scalable agentic workflows using LangGraph, PydanticAI and MCP for modular orchestration. His GitHub contributions to Keras model implementations show attention to robustness and test-driven fixes, reflecting a pragmatic balance of research, engineering, and product delivery.
11 years of coding experience
6 years of employment as a software developer
Master's degree, Computer Science and Engineering, Master's degree, Computer Science and Engineering at Politecnico di MilanoBachelor's degree, Electrical, Electronics and Communications Engineering, Bachelor's degree, Electrical, Electronics and Communications Engineering at Istanbul Technical University
English
Github Skills (5)
keras10
machine-learning10
deep-learning10
tensorflow10
python10
Programming languages (7)
Github contributions (5)
rcmalli/keras-vggface
Oct 2016 - Mar 2020
VGGFace implementation with Keras Framework
Role in this project:
ML Engineer
ML EngineerContributions:2 releases, 1 review, 76 commits in 3 years 5 months
Contributions summary:Refik primarily contributed to the `keras-vggface` repository, which implements VGGFace models using the Keras framework. Their work focused on addressing issues related to model functionality, including fixing download links for model weights and correcting layer names. They also added support for a newer version of Keras, and refactored the code. They also added unit tests to confirm the models correctness.
deep-learningkerasvggfacetensorflow
rcmalli/keras-squeezenet
Oct 2016 - Feb 2019
SqueezeNet implementation with Keras Framework
Role in this project:ML Engineer
ML EngineerContributions:1 release, 42 commits, 6 PRs in 2 years 4 months
Contributions summary:Refik primarily contributed to the implementation and maintenance of a SqueezeNet model using the Keras framework. Contributions include renaming model files, releasing new versions, fixing setup issues, and preparing for new version releases. The user also updated the test script to ensure the model's functionality.
deep-learningkerassqueezenettensorflow
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.