I am an applied AI engineer and research professional focused on building reliable machine learning systems, particularly in computer vision and modern language-based architectures (LLMs, RAG, and agents).
Over the past years, I have worked on projects that combine deep learning with pragmatic system design-covering model development and evaluation, data strategy (including synthetic data when appropriate), and deployment constraints such as latency, resource usage, and maintainability.
I typically work across the full lifecycle: defining the problem and KPIs, designing modular pipelines, training and validating models, and operationalizing them with reproducible environments and experiment tracking.
My current research areas of interest are timeseries modeling, computer vision and language systems (LLMs). I also have a background in algebra and cryptography since I did some work in Criptografia i Grafs (CiG) research group in Universitat de Lleida.
Publications
Peer-reviewed journal articles
Alàs, O., Cervera, R., & Sanfeliu, R. (2026). A Scalable Real-Time Multi-Camera Vehicle Tracking System for Urban Environments. IEEE Access. doi: 10.1109/ACCESS.2026.3675896
Velasquez-Camacho, L., van Doorn, N., Preisler, H., Etxegarai, M., Alas, O., Gonzalez Castro, J. M., & de-Miguel, S. ( 2025). Monitoring temporal changes in large urban street trees using remote sensing and deep learning. PLOS ONE, 20( 6), e0326562. doi: 10.1371/journal.pone.0326562
Alàs, O., & Sebé, F. (2024). Privacy-Preserving Electricity Trading for Connected Microgrids. Applied Sciences, 14(4), 1458. doi: 10.3390/app14041458
Thesis / academic work
Alàs Cercós, O. (2023). Generative deep learning based models for cloud removal satellite imagery (Master’s thesis / Treball Final de Màster, Universitat de Lleida – EPS; public defense dated 27 Sep 2023). Link
Alàs Cercós, O. (2021). Implementació d’un sistema criptogràfic per l’enviament del consum elèctric en sistemes de comptadors intel·ligents (Treball Final de Grau, Grau en Enginyeria Informàtica, Universitat de Lleida – Escola Politècnica Superior; presented July 2021; supervisor: Francesc Sebé Feixas). Link
Hackathons
I participated in hackathons as a way to explore new ideas and rapidly prototype systems.
These projects typically involve building functional prototypes within 24–72 hours.
- MediSur. AI-assisted medical triage and appointment management platform designed to optimize healthcare workflows.
- Spinning Enjoyers (Winner bitsxlamarató 2022). Wearable health monitoring system that detects potential stroke symptoms using sensor data from a smartwatch.
- Sales Prediction RNN (Winner HackUPC 2022). Time-series forecasting system that uses recurrent neural networks to predict product sales based on historical demand data.
- JobCall (Winner HackUPC 2021). Intelligent job-matching platform based on speed datings that connects candidates with relevant opportunities using data-driven recommendation mechanisms.
Hackathons are where I explore new ideas in AI, rapid prototyping architectures that later evolve into more complete systems.
More projects on Devpost.