Aboutme

Aboutme

Areas of interests

  • Machine Learning
  • Deep Learning
  • Deep Reinforcement Learning

Resume (English)  |  Resume (French)

Experience

Applied AI Researcher

Conservatoire National des Arts et Métiers (Cnam) – Paris | 2023 – Present

  • TeleSLMs: Full fine-tuning of SmolLM2 (135M, 360M) on 1.87B tokens from 38,302 3GPP/ETSI standards documents, followed by LoRA instruction fine-tuning. SmolLM-TS-135M-it achieves 31.5% Ans-PPL reduction (SemScore 0.6183→0.6504); SmolLM-TS-360M-it achieves 20.6% reduction (SemScore 0.6216→0.6572) vs baseline on Tele-Eval. 4 models + TeleSpec-Data dataset published on HuggingFace. Multi-GPU training via PyTorch DDP on 3× L40S.
  • NEXASPHERE Project: Design AI framework for automatic anomaly detection and recovery in 3D networks (terrestrial/satellite).
  • INFLUENCE Project: Developed the “Autonomous Reconfiguration – Intent Management” architecture with a DRL agent for intelligent container scaling.
  • 5G3E Dataset: Led the creation of 5G3E — an open-source end-to-end 5G telemetry dataset for training telecom AI models.
  • AI Strategy: Designed full technical architecture and ML innovation roadmap for the proposal submissions to the national innovation programs (i-Lab Innovation Competition, France 2030 – Pionnier de l’IA).
  • Funding propsals: Contributed to the development of three EU research proposals, one of which was awarded funding (NEXASPHERE).
  • AI4CI European Program: Designed and delivered the advanced module “AI & ML for Connected Systems” for a European Master program.

PhD Researcher

Avignon University | ANR MAESTRO5G Project | 2019 – 2022

  • Developed mathematical models and AI algorithms for dynamic resource allocation in 5G networks.
  • Applied Markov Decision Processes (MDP) to orchestrate IoT data flows, optimizing Age of Information (AoI) and reducing operational network infrastructure costs.

Intern

Azcom Technology, Milan, Italy | 2018 – 2019

  • Developed a machine learning framework based on the KNN algorithm for detecting false peaks in 4G mobile communications, later adapted for 5G networks.

Education

PhD in Computer Science — Avignon University, France2022
Telecommunications Engineering Degree — Politecnico di Milano, Italy2019

Patent & Publications

Patent: Methods for detecting and managing anomalies impacting a computing environment, corresponding devices and computer programs, 2024 — Ref: FR3164046.

Journal papers

  1. Joint Traffic Offloading and Aging Control in 5G IoT Networks — Naresh Modina, Rachid El Azouzi, Francesco De Pellegrini, Daniel Sadoc Menasche, Rosa Figueiredo. IEEE Transactions on Mobile Computing.

Conference papers

  1. Naresh Modina, Riccardo Ferrari, Maurizio Magarini — “A machine learning-based design of PRACH receiver in 5G”. Procedia Computer Science 151, 1100-1107.
  2. Naresh Modina, Rachid El-Azouzi, Francesco De Pelligrini, Daniel Sadoc Menasche — “Joint traffic offloading and aging control in 5G IOT networks”. 2020 32nd International Teletraffic Congress (ITC 32), 147-155.
  3. Naresh Modina, Mandar Datar, Rachid El-Azouzi, Francesco De Pelligrini — “Multi-resource allocation for network slices with multi-end fairness”. 2022 IEEE International Conference on Communications (ICC 2022).
  4. Salah Ali Bin Ruba, Naresh Modina, Pedro Braconnot Velloso, Stefano Secci — “FLADxG: Federated Learning Based Anomaly Detection Framework for xG Systems”. 2024 IEEE 13th International Conference on Cloud Networking (CloudNet).

Talks

Technical Skills

Python

  • Data & MLOps: NumPy, Pandas, Docker, Git
  • Machine Learning: Scikit-learn
  • Deep Learning: TensorFlow, PyTorch — CNN, RNN, LSTM, Transformers
  • Deep Reinforcement Learning: Gymnasium — DQN, DDQN, D3QN, PPO, World Models
  • Visualization: Matplotlib

Document Writing

  • LaTeX
  • Markdown