About
I am an Electrical engineering and Master in signal and image processing with hands-on experience designing and evaluating deep learning algorithms and Gaussian Processes for medical imaging, time-series analysis, and model optimization. I am motivated by the challenge of designing high-quality and reliable machine learning models for real-world data, and I look forward to continuing to learn and contribute to high-impact research as I grow both professionally and academically.
Download CV
Research Interests
- Deep Learning Algorithms
- Gaussian Processes
- Medical Imaging
- Computer Vision
- Image & Signal Processing
- Time-Series Analysis
- Model Optimization
- Uncertainty Quantification
Publications
- L. F. Torres Torres, S. F. Osorio-Botero, X. Palard, R. De Crevoisier, L. Cubero, O. Acosta, J. Castelli, M. Rubeaux. "Inclusion of Ipsilateral Salivary Gland PET/CT Features Improves Xerostomia Prediction After Radiotherapy in Head and Neck Cancer." Highlight virtual poster, ESTRO 2026, Stockholm, Sweden.
- S. A. Cajas Ordóñez, L. F. Torres Torres, M. J. Meni, C. A. Duran Paredes, E. Arazo, C. Bosch, R. S. Carbajo, Y. Lai, L. A. Celi. "Uncertainty Makes It Stable: Curiosity-Driven Quantized Mixture-of-Experts." arXiv preprint arXiv:2511.11743, 2026 [link].
- L. F. Torres-Torres, J. Arias-García, H. F. García, A. F. López-Lopera, J. F. Vargas-Bonilla. "Gaussian Process-driven Hidden Markov Models for Early Diagnosis of Infant Gait Anomalies." IEEE EMBC 2025, Copenhagen, Denmark [DOI].
- S. A. Cajas Ordóñez, L. F. Torres Torres, M. Bifulco, C. A. Durán, C. Bosch, R. S. Carbajo. "Embedding-Aware Quantum-Classical SVMs for Scalable Quantum Machine Learning." arXiv preprint arXiv:2508.00024, 2025 [link].
- V. Aher, S. A. Cajas Ordóñez, L. García, E. Salas Villa, L. Torres, V. K. Verma. "Intensity-Based Prompt Generation for Multi-Modality 3D Medical Image Segmentation." CVPR 2025 Workshop (BiomedSegFM Challenge).
- H. F. García Arias, L. García-Mosquera, L. F. Torres-Torres, A. Passo, I. Uribe. "Translational Telemedicine for Indigenous Health: A Community-Centered Approach in the Karmata Rúa Reservation." 2025 IEEE Global Humanitarian Technology Conference (GHTC), 2025.
Experience
Research Experience
Research Intern — Salivary Reserve in ¹⁸F-FDG PET Imaging
2025 – Current
Measurement of the Salivary Reserve in Head and Neck Cancer | Laboratoire Traitement du Signal et de l'Image (IMPACT team), Université de Rennes
- Developed a method for the automatic detection of outliers to identify tumor invasion in salivary gland segmentations using Bayesian bootstrap probability aggregation to measure epistemic uncertainty across iterations on the ARTIX dataset (CEM).
- Built an automatic voxel-based model to clean tumor apparitions from segmentations using imaging features (intensity, local neighborhood, texture/radiomics, gradient/boundary, probabilistic, and geometric/spatial).
- Predicted xerostomia based on SUV information extracted from PET imaging and compared performance against the standard clinical assessment protocol used at Centre Eugène Marquis.
Co-authored Publication — Curiosity-Driven Quantized Mixture-of-Experts
2026
"Uncertainty Makes It Stable: Curiosity-Driven Quantized Mixture-of-Experts" | Université de Rennes
- Designed a curiosity-driven Bayesian routing algorithm for Mixture-of-Experts, using Monte Carlo dropout and KL divergence for adaptive, uncertainty-aware expert selection among heterogeneous quantized models.
- Achieved a 50% to 94% reduction in F1 variance (p < 0.001) through interpretable, confidence-based expert selection, in which uncertain samples are automatically referred to more accurate models.
- Demonstrated 4-bit quantization preserving 99.9% of full-precision F1 (0.858 vs 0.859) with 4x compression and 31% energy savings, targeting responsible and sustainable edge deployment.
Undergraduate Thesis — Gaussian Process-driven Hidden Markov Models
2025
"Gaussian Process-driven Hidden Markov Models for Early Diagnosis of Infant Gait Anomalies" | Machine Learning & Robotics Group (UdeA)
- Extracted biomechanical vectors of key points to represent joint movements during Gait Cycles using pre-trained pose estimation models (OpenPose, Pose2Sim, Halpe26).
- Designed and trained multi-output Gaussian Processes to capture the temporal dynamics of gait and the interdependencies of gait signals, obtaining metrics of MAE = 1.659, R2 = 0.963, aDTW = 34.920.
- Implemented HMM-based gait phase segmentation for anomaly detection in infant gait temporal patterns, validated on clinical data.
Electronics engineering, Computer vision and AI modeling
August 2022 – June 2023
Young research intern in ALIES: Alimentadora de esquejes | GEPAR UdeA [Repository]
- Programming of the Computer Vision Algorithm for feature extraction to select and classify the types of cuttings (flowers) using the OpenCV library in Python.
- Calibration of the Ueye-IDS camera, with all its environment to obtain images in real time, and process the data for statistical analysis.
- Development of a desktop software with a graphical interface that displays statistics of the images stored in a database, provides information to the user and controls the operation of the machine
- Establish the integration between the software and hardware system for the precision agriculture system.
Education
Master 2 SiVOS (Signal, Image, Vision, Optronics & Systems)
Current
Université de Rennes, France
Electronic Engineering
2019 - 2025
Universidad de Antioquia, Medellin
Técnico en programación de software
2014 - 2016
Servicio Nacional de Aprendizaje (SENA)
Volunteering & Leadership
Student Chair — Telemedicine Baseline Project
2025
IEEE Humanitarian Engineering Program, Colombia
Co-founder — EMBC Student Chapter
2025
Universidad de Antioquia
President — IEEE Student Branch
2024
Universidad de Antioquia
Leader & Co-founder — Machine Learning & Robotics UdeA Student Group
2022 - 2024
Universidad de Antioquia
Vicechair — Aeronautics and Drones Division
2021 - 2023
IEEE-AESS Unicauca
- Researcher in aerospace electronics, space robotics, drones, image processing and AI; contributed to the DSTEI drone-swarm project for coffee-crop surveillance using CNNs and multispectral imaging.
- 1st place ALA ZAGI Race 2022 (AI / autonomous driving) and team leader of the ROVEARTH project at the IEEE Radar Challenge 2022 in New York [Link].
- Delivered STEM workshops for youth and co-organized outreach events such as Exploring the Future and SPACE WEEK.
Grant/Awards
- Distributed Sensing Technology and Education Initiative (DSTEI) IEEE AESS Scholarship ($25.000): Grant to finance the design phase of a drone swarm system for coffee crop monitoring. The design and education process in aerospace systems engineering is conducted in 4 countries over a period of one year [link]
- Mentoring Experiences for Underrepresented Young Researchers (ME-UYR) IEEE SPS ($4.000) to attend ICASSP [link]: Financial support to cover travel and accommodation expenses to present a research paper at the ICASSP 2025 conference with University of Alberta mentor Dr. Li Cheng.
- Tech 4 Good Challenge Colombia ($10.000): Winner with IEEE to develop a baseline telemedicine project in the Karmata Rúa indigenous reservation, establishing primary care stations.
- Google Cloud Research Credits Program ($5.000): GCP grant under award number GCP19980904 supporting Quantum Computing and Machine Learning model optimization projects.
- CVPR 2025 Challenge — 4th place in Foundation Models for Interactive 3D Biomedical Image Segmentation with the proposed MobileSeg3D framework, a lightweight solution for multi-modality 3D medical image segmentation.