Pol Caselles Rico Email & Phone Number

Barcelona, Cataluña, España

Working on research, design, and implementation of Computer Vision algorithms with Deep Learning for 3D reconstruction. During my PhD, some of the work done has been published in conferences (Industrial Doctorate fellowship. Supervisors: Francesc Moreno and Gil Triginer).I have been working as an individual contributor in launching the following products to the market: High-quality face reconstruction and its lightweight version. Fast full body and head reconstruction from a few input images. These developments directly impacted the retention and adoption of the core product of the company serving thousands of customers. Hands-on experience in software development and maintenance of data pipelines, model life-cycle, and end-to-end development from research to production. Designed and developed production-ready deep-learning models such as but not limited for landmarks prediction, normals and depth estimation, shape estimation (neural fields), in-painting with diffusion, and implementation of state-of-the-art papers in 3D reconstruction. Proficiency in Deep Learning frameworks and libraries like PyTorch, Torchvision, OpenCV and strong mathematical skills. Familiarity with Python, Git, Docker, Wandb and MLflow.

Barcelona, Cataluña, España

Invited talks about 3D Reconstruction:- UPC School- Barcelona Technology School (BTS)Lecturer at the Artificial Intelligence with Deep Learning postgraduate course:- Deep Learning and Computer Vision subjects.- Project advisor. e.g. Yoga pose estimation, Image Colorization with GAN'sTopics covered: - Generative models (GANs, VAEs, Diffusion)- NLP (RNN, GRU, LSTMs, Transformers)- Optimization, Transfer learning, distillation, model training, regularisation- Object detection and instance segmentation- Reinforcement Learning (Tabular Q, DQN, PPO)

Tübingen, Baden-Württemberg, Germany

I collaborated with the Real Virtual Humans group under the supervision of Dr. Gerard Pons-Moll.During my stay, we worked on human mesh recovery from single-input images. We conducted hands-on research on diffusion models for point cloud generation.

Barcelona, Cataluña / Catalunya, España

Developing an API for computing ESG Rating online. (AWS, Python)- Managing user data, forms, computations and results.- Backend: MongoAtlas (MONGO M1), AWS (Lambda, ECR, API REST)Data analytics for consumer Insights (Obra Social "la Caixa", Fundació Puigvert, Suez Water Spain)- Customer clusterization, (DBSCAN, PCA, T-SNE/UMAP)

Sankt Gallen, Suiza

Master's Thesis Topic: Learning representations of the network's weight space.https://upcommons.upc.edu/handle/2117/345453?show=fullUse transformer architecture to investigate the interaction of the model’s weight during its training.

Barcelona Y Alrededores, España

- Worked on computer vision algorithms using Deep learning for automated depalletizing system.- Designing database's structure for AI in production environments and implementation of an inference server (MongoDB)- Development of a machine learning platform that enables automatic annotation of datasets (C#)

County Dublin, Ireland

Bachelors Thesis: Integrating low-level motion cues in deep video saliency, AI Project, computer vision.Supervisors: Kevin Mcguinness, Xavier Giró Nietohttps://upcommons.upc.edu/handle/2117/168654- Exploit optical flow information to enhance video saliency prediction (using deep learning).- I took a classical machine learning course (RF, KNN, SVM/SVR.., )

Doctor Of Philosophy - Phd, Computer Science

IRI - Institut de Robòtica i Informàtica Industrial, CSIC-UPC. Supervisors: Francesc Moreno Noguer and Gil Triginer (from Crisalix)..

Master'S Degree In Advanced Telecomunications (Matt), Artificial Intelligence

Deep learning for multimedia processing: - Computer vision - Speech and language processing - Deep Learning for Artificial Intelligence

Bachelor'S Degree, Telecommunications Engineering