Selected work
Projects
Research and engineering across computer vision, multimodal learning, and applied AI.
QueriSight — Multimodal Video Retrieval
Designing a multimodal video retrieval pipeline that transforms videos into searchable, timestamped segments using scene-aware segmentation, keyframe extraction, SigLIP visual embeddings, and Qwen-generated descriptions.
IPCV final project · Sep 2026 – Present · Supervisor: Fabien Baldacci
Supervisor profile ↗Cross Modal Attention Bridge
Investigating CLIP interpretability through attention rollout, token filtering, and perturbation-based XAI methods including LIME, SHAP, and RISE.
TRDP project · Nov 2025 – Present · Supervisor: Dr. Marcos Escudero-Viñolo
Individual HRV Response to Music via PPG
Designed a within-subjects PPG HRV study on music-induced autonomic modulation with Butterworth filtering, IBI extraction, FFT-PSD, and non-parametric statistical analysis.
Sep 2025 – Dec 2025 · Supervisor: Dr. Márton Áron Goda
Supervisor profile ↗Weighted Embeddings Fusion for Hate Detection
Designed an SE-based multimodal fusion model using CLIP, TweetEval, and EfficientNetB3 for meme classification, achieving SOTA performance with an F1 score of 92.5% on the MultiOFF and MAMI datasets.
Sep 2024 – Jan 2025 · Supervisor: Dr. Ram Sarkar
Supervisor profile ↗