Alejandro Golfe
AI Portfolio
AI Engineer
About Me
Industrial engineer specialized in artificial intelligence. Quick learner, adaptable, and goal-oriented professional committed to delivering results responsibly and efficiently.
Education
-
PhD in Health and Wellness Technologies (AI Focus) — Polytechnic University of Valencia (UPV)
2022 – Present -
Master’s Degree in Artificial Intelligence — Valencian International University (VIU)
2021 – 2022 -
Master’s Degree in Industrial Engineering — Polytechnic University of Valencia (UPV)
2018 – 2021 -
Bachelor’s Degree in Industrial Technologies Engineering — Polytechnic University of Valencia (UPV)
2014 – 2018
Work Experience
AI Engineer — CVBLAB (UPV)
2021 – Present
- Conducted research and trained state-of-the-art generative models (GANs, Diffusion Models) for high-fidelity image generation.
- Built CBIR systems, improving image search accuracy and speed.
- Processed and cleaned large multimodal datasets for model training.
- Optimized training pipelines, reducing GPU/TPU time.
- Fine-tuned pre-trained models for vision and multimodal tasks.
- Evaluated models using quality, diversity, and relevance metrics.
- Led cross-functional teams in AI projects, coordinating meetings, timelines, and deliverables.
- Built and optimized deep learning models for medical image analysis to support cancer detection.
Projects
BlastDiffusion: A Latent Diffusion Model for Generating Synthetic Embryo Images to Address Data Scarcity in In Vitro Fertilization
| Code | Publication |
Accurately identifying oocytes that progress to the blastocyst stage is crucial in reproductive medicine.
We propose BlastDiffusion, a generative model based on Latent Diffusion Models (LDMs), which synthesizes realistic oocyte images conditioned on developmental outcomes, addressing the challenges of limited annotated embryo datasets.
Key Highlights
- Utilizes a pretrained Variational Autoencoder (VAE) for latent space representation combined with a diffusion process to generate realistic images.
- Produces synthetic images that differentiate oocytes that reach the blastocyst stage from those that do not.
- Achieves superior performance compared to GAN-based models
- Qualitative analysis shows the model captures key morphological differences linked to developmental outcomes.
- Demonstrates the potential of diffusion models for data augmentation and automated embryo assessment in IVF.
Enhancing Image Retrieval Performance with Generative Models in Siamese Networks
This research addresses the challenges of prostate cancer diagnosis, one of the most prevalent cancers among men worldwide.
We propose a Siamese Network approach to encode image patches into latent representations for Content-Based Image Retrieval (CBIR) tasks, leveraging generative deep learning models to improve retrieval quality.
Key Highlights
- Leveraging the ProGleason-GAN framework trained on the SiCAPv2 dataset for generating synthetic patches.
- Achieved notable improvements in retrieval metrics with synthetic data augmentation.
- First approach where CBIR-optimized latent representations are used to train an attention mechanism for Gleason Scoring of Whole Slide Images (WSI).
ProGleason-GAN: Conditional Progressive Growing GAN for Prostatic Cancer Gleason Grade Patch Synthesis
| Code | Publication |
This research focuses on prostate cancer, one of the most common diseases affecting men worldwide.
We propose ProGleason-GAN, a conditional Progressive Growing GAN designed to synthesize histopathological patches of prostate tissue for data augmentation, addressing the challenges of insufficient and unbalanced datasets in Gleason grading models.
Key Highlights
- Conditional GAN architecture capable of generating patches for non-cancerous patterns, GG3, GG4, and GG5 by embedding Gleason grade information.
- External validation performed by expert pathologists confirmed the realism of the synthetic samples.
- Significant improvement in classification performance on the SiCAPv2 dataset with synthetic data augmentation.
Skills
Technical Skills
- Programming & Frameworks: Python, PyTorch, TensorFlow, Hugging Face, OpenCV, NumPy, Pandas
- Generative AI: GANs, Diffusion Models, Transformers
- Computer Vision & CBIR: Image retrieval, object detection, segmentation, multimodal data processing
- ML Engineering & Pipelines: Data preprocessing, model training, hyperparameter tuning, GPU/TPU optimization
- Research & Evaluation: Model metrics, fine-tuning, transfer learning
- Tools & Deployment: Git, Docker, Colab, Hugging Face Spaces, basic Streamlit
Leadership & Product Skills
- Project management, team coordination, roadmap planning
- Translating research into deployable AI solutions
- Cross-functional team leadership and stakeholder communication
Languages
- English: Proficient
- Valencian: Native
- Spanish: Native
Contact
- 📧 Email: alejandrogolfe@gmail.com
- 📱 Phone: +34 627 245 209
- 📍 Location: Valencia, Spain