AI Challenge 2025 Winners

[robertogl]

AI Challenge Winners

Artificial intelligence is reshaping how we tackle complex problems across disciplines. In this spirit, our AI challenge drew a remarkable range of creative, technically rigorous submissions from around the world, making judging especially difficult. After careful deliberation, we recognized three standout projects for their advanced uses of Artificial Intelligence in physics, signal processing, and fluid dynamics applications. Each offered a distinctive approach with practical implications and real-world validation. We are pleased to highlight these top entries and congratulate all participants for their creativity and dedication!

---

Winners at a Glance

Place	Winner	Project	Domain	Highlight
🥇	Adit Shah	Top Quark Detection with Deep Learning and Big Data	High-Energy Physics	>90% test accuracy; ResNeXt + SE
🥈	Nicola Gallucci, Matteo Malagrino & Giacomo Aragnetti	Classify RF Signals Using AI	Signal Processing	Pixel-level spectrum maps; SDR-based validation
🥉	Soham Gupta	Fluid Flow Simulation Using Physics-Informed Neural Networks	Fluid Dynamics	Physics-consistent fields; viscosity estimation

---

🥇 1st Place

Adit Shah

SVKM's Dwarkadas J. Sanghvi College of Engineering, India

Challenge project completed: Top Quark Detection with Deep Learning and Big Data

Important

Why it won: Clear physics grounding, strong accuracy, innovative method based on ResNeXt blocks with Squeeze-and-Excite (SE) attention and a credible hardware-deployment path.

Tip

Standout results: Aggregated residual (ResNeXt-style) blocks + Squeeze-and-Excite; robust preprocessing; HDL/FPGA workflow outlined.

In high-energy physics, particle collisions leave “sprays” of energy called jets. Adit’s solution turns raw jet data into multi-channel images and uses a custom CNN, designed with the Deep Network Designer app from the Deep Learning Toolbox™, to distinguish top-quark events from background. The approach pairs thoughtful preprocessing with aggregated residual (ResNeXt-style) blocks and Squeeze-and-Excite attention so the model focuses on real physics patterns, not noise, achieving over 90% test accuracy. It also outlines a workflow for HDL code generation and FPGA deployment, bridging lab research to real-time systems.

This work highlights Adit’s innovative thinking and an end-to-end engineering mindset, reframing a physics problem as an image task, selecting modern model architectures, and planning for hardware deployment. Congratulations to Adit on an outstanding first-place achievement!

Tech stack: MATLAB®, Deep Learning Toolbox™ (Deep Network Designer app), HDL code generation & FPGA deployment workflow.

---

🥈 2nd Place

Nicola Gallucci, Matteo Malagrino, Giacomo Aragnetti

Polytechnic of Milano

Challenge project completed: Classify RF Signals Using AI

Important

Why it won: End-to-end system thinking—careful dataset design, strong model choices, and decisive real-world validation.

Tip

Standout results: Pixel-level maps of Wi-Fi, Bluetooth, ZigBee, SmartBAN from STFT spectrograms; two synchronized ADALM-Pluto SDRs capture and stitch 80 MHz of spectrum.

Nicola and Matteo tackle the crowded 2.4 GHz band by turning STFT spectrograms into pixel-level maps of Wi-Fi, Bluetooth, ZigBee, and SmartBAN activity. Their complete pipeline trains attention-gated U-Net and DeepLabv3+ (with ResNet-18/50 encoders) on a synthetic, channel-aware dataset with class-priority masks, then validates on real airwaves using two synchronized ADALM-Pluto SDRs to capture and stitch 80 MHz of spectrum. Built in MATLAB® with Computer Vision Toolbox™ and Parallel Computing Toolbox™, and using Deep Learning Toolbox™ and Signal Processing Toolbox™ workflows, the solution is ready for practical spectrum monitoring and interference mitigation.

This work highlights Nicola, Matteo, and Giacomo’s innovative, end-to-end engineering, from careful dataset design and model choice to decisive real-world validation. Congratulations to Nicola, Matteo, and Giacomo on a fantastic achievement!

Tech stack: MATLAB®, Deep Learning Toolbox™, Computer Vision Toolbox™, Signal Processing Toolbox™, Parallel Computing Toolbox™, Two synchronized ADALM-Pluto SDRs for real-world capture.

---

🥉 3rd Place

Soham Gupta

Indian Institute of Technology, Ropar

Challenge project completed: Fluid Flow Simulation Using Physics-Informed Neural Networks

Important

Why it won: Strong scientific rigor with an approachable, reproducible workflow that others can extend.

Tip

Standout results: Physics-consistent 2-D fields around a cylinder; estimates viscosity; clear comparisons of predicted vs. true velocity, pressure, and vorticity.

Soham’s project builds a physics-informed neural network in MATLAB® to solve 2-D incompressible flow around a cylinder. Instead of learning from labeled examples, the network is trained to obey the governing equations and boundary conditions, so its predictions stay physically consistent. It produces the full flow field—velocity (u, v) and pressure (p)—and even estimates the fluid’s viscosity. A simple two-step training recipe and well-organized scripts make the work easy to reproduce, with clear visuals comparing predicted and true velocity, pressure, and flow rotation (vorticity).

This work highlights Soham’s blend of scientific rigor and practical engineering, turning core fluid-mechanics principles into a reliable learning system that others can run and extend. Congratulations to Soham on an elegant and insightful achievement!

Tech stack: MATLAB®, Custom PINN training and visualization scripts.