Advanced Machine Learning Researcher
Passionate about turning cutting-edge AI concepts into impactful solutions. Specialized in machine learning, data science, and innovative technology applications.
Get to know me better
Dr. Arijit Nandi is an Advanced AI researcher and engineer with a Ph.D. in Artificial Intelligence from UPC-BarcelonaTech and over 6+ years of experience building scalable, privacy-preserving, and interpretable AI systems for real-world deployment. His work spans federated learning for privacy-preserving model development, explainable AI (XAI) for transparency and accountability, and applied deep learning across time-series, sensor data, and language-based systems. On the applied side, he builds LLM-powered tools for intelligent automation — including content personalization, workflow optimization, and decision-support assistants — alongside an XAI-as-a-Service platform aimed at democratizing model interpretability. He architected EDFL, an open data space federated learning framework recognized by the European Commission on its Innovation Radar.
He bridges academic rigor and production deployment, with experience across the full ML lifecycle: from research and prototyping to containerized, scalable systems in real-world environments. His focus is on making AI not just performant — but transparent, safe, and human-centric.
Technologies and tools I work with
My professional journey
Leading AI initiatives and developing machine learning solutions for enterprise clients.
Contributed as an AI expert to design and develop emotion recognition from facial expression using Tensorflow 2.0 and FER 2013, CK++. The trained model is deployed in the backend of Tuttify (https://tuttify.io/) to recognize different emotions of the students.
Contributed as an AI expert to design and develop AI MOOC course for young people (AIM4YOU) under the EU project YNSPEED (Youth new personal & employable skills development).
Some of my recent work
Designed and deployed a Docker-enabled federated learning framework (DFL) for multi-modal data stream classification. Clients and global servers communicate via lightweight MQTT protocol, enabling real-time emotion state classification from distributed physiological data (EDA + RB) while preserving data privacy. Published in Computing (Springer, 2023).
Developed a novel Reward-Penalty Based Weighted Ensemble (RPWE) approach for emotion state classification from multi-modal physiological data streams (DEAP & AMIGOS datasets). Classifiers are dynamically rewarded or penalized based on predictive performance, with an auto-adjusting beta factor. Published in International Journal of Neural Systems (2022).
Built a federated learning framework (Fed-ReMECS) for real-time emotion state classification from multi-modal physiological data streams via wearable sensors. Uses MQTT for IoT communication and builds a global classifier without accessing users' local data, ensuring privacy. Evaluated on the DEAP dataset. Published in Methods (Elsevier, 2022).
My academic background
Specialized in Artificial Intelligence and Machine Learning with focus on deep learning applications.
Specialized in Artificial Intelligence and Machine Learning with focus on deep learning applications.
Graduated with honors. Focused on software engineering and data structures.
Research contributions and scientific work
Loading publications...
Let's work together
arijit4ai@gmail.com
+34 (if you have!!)
Connect With Me
Follow my research and professional journey