Skills
Machine learning & deep learning (Pytorch) | Generative models: VAEs, GANs, Normalizing Flows, Diffusion | LLMs | 3D data processing & motion capture analysis | Scientific dissemination | Cross-functional collaboration in interdisciplinary research groups | Languages: Spanish, English, German, French | Python, C++, Numpy, OpenCV
Experience
Fraunhofer HHI, Germany. Research Associate (January 2021 – Present)
As a research associate at the Computer Vision and Graphics group, I conduct research in generative AI (genAI), with a focus on virtual avatar animation for extended reality (XR), including applications in virtual training simulations, teleconferencing, and interactive storytelling. I collaborate cross-functionally with academic and industry partners on large-scale projects involving large language models (LLMs), generative models, and 3D human motion synthesis (e.g. LUMINOUS, SPIRIT, MoDL, INVICTUS). My work involves designing and implementing neural network architectures for realistic motion generation in avatar-based systems. I actively track developments in generative AI (e.g. diffusion models, transformers, GANs), integrating research advances into research workflows. I supervise students involved in ongoing research activities, including one current Master’s thesis under my guidance on transformer-based gesture synthesis for avatar animation. I disseminate research findings through internal seminars, partner meetings, and external conference workshops (e.g. Web3d, CVMP, EuroXR). I support project reporting, documentation, and technical communication with non-technical stakeholders.
Achievements:
- Co-authored a peer-reviewed paper on generative human motion synthesis, awarded Runner-Up Best Paper at CVMP 2024 for innovation in controlled motion generation.
- Designed and benchmarked a GAN-based model for avatar animation from limited motion data captured in our studio, achieving 99-100% training coverage and increased global/local motion diversity with enhanced controllability.
- Integrated feet contact modeling and motion constraints to improve physical realism and motion continuity in synthesized human avatars.
- Collaborated with 15+ academic and industry partners across AI, HCI, XR, healthcare, and training sectors in EU-funded multi-institutional research initiatives.
- Presented research outcomes at EuroXR, Web3D, and CVMP, and led technical dissemination to partners and stakeholders.
- Contributed to securing continued funding through high-impact grant deliverables, live research demos, and technical documentation.
- Improved model generalization through custom data preprocessing and augmentation pipelines.
Vicomtech, Spain. Computer Vision Intern (February 2020 – June 2020)
Conducted applied research in computer vision and real-time generative systems for artistic performance contexts. Developed a working prototype that captured audience body movements using a multi-person pose estimation model, which guided dynamic 3D effects and modulated the latent space of a pre-trained GAN to generate real-time stylized visuals. Extracted musical features to further drive synthesis and sync animations with live sound. Integrated a face motion transfer model as an experimental extension for additional avatar animation capabilities. Worked under the supervision of the Digital Media Department and in collaboration with a cultural venue partner.
Achievements:
- Retrained StyleGAN on custom datasets to offer specific visual styles and animations for live artistic output.
- Optimized and repurposed open-source models for real-time performance constraints, maintaining frames of ~15-30 FPS.
- Presented a live prototype to a partner venue for potential deployment in public installations.
- Delivered a fully functional prototype demonstrating multimodal interaction (pose and audio) for artistic generation.
Université de Bordeaux, France. Student Researcher (January 2019 – December 2019)
Conducted research on the challenges of object detection in immersive image formats, focusing on the effects of spherical and equirectangular projections on model performance. Analyzed how geometric distortions introduced by panoramic projections impact detection accuracy and spatial consistency, and methods to mitigate them. Reviewed state-of-the-art deep learning methods for object detection (e.g., YOLO, Faster R-CNN) and evaluated their applicability to 360° image data.
VASS LATAM (Client: AXA), Colombia. Consultant – Software Developer (October 2017 – August 2018)
Developed core software components for AXA within a highly regulated enterprise environment, following strict documentation, versioning, and quality assurance protocols. Contributed to system integration workflows using Microsoft BizTalk Server and supported backend development tasks within the .NET ecosystem. Collaborated with a consulting team to deliver client-facing solutions aligned with AXA’s internal software standards.
Pontificia Universidad Javeriana, Colombia. Student Researcher (January 2016 – July 2016)
Conducted research on image quality assessment methods for fused infrared and visible spectrum imagery. Investigated natural scene statistics in multi-spectral image data and developed models that better approximate human perceptual quality. Work involved statistical feature extraction, perceptual modeling, and benchmarking against existing image fusion evaluation techniques.
Achievements:
- Developed two image quality evaluation models that outperformed existing metrics in correlation with human subjective ratings, achieving improvements of +0.14 absolute (~25% relative) and +0.16 absolute (~20% relative) when compared to the best performing baselines.
- Presented research at ICASSP 2017 and published findings in IEEE Transactions on Image Processing (TIP).
- Demonstrated the application of perceptual modeling to multi-spectral image fusion in a cross-modal context.
Continental Automotive GmbH, Germany. R&D Intern (April 2015 – July 2015)
Contributed to the development of a test reporting tool to analyze results from electronic control units (ECUs). Enhanced reporting features using C# and the .NET framework to generate Excel-based statistics, enabling deeper insights into test performance. Gained experience in automotive testing workflows, object-oriented programming, and software development practices in a multicultural R&D environment.
Education
- Université de Bordeaux, Universidad Autónoma de Madrid, Pázmany Péter Katolikus Egyetem - MSc Image Processing and Computer Vision (2018 - 2020): Fully funded Erasmus+ student.
- Pontificia Universidad Javeriana de Cali - BSc in Electronics Engineering (2010-2017): Honor distinction for Thesis Natural Scene Statistics of Long Wave Infrared and Visible Images, Merit awards.
- Karslruhe Institute of Technology (2015): Exchange semester, DAAD scholarship.
Languages
| Language | Level | Certificate |
|---|---|---|
| Spanish | Native | |
| English | C1 | IELTS - 7.5 |
| German | B2 | TestDaF |
| French | A2 |
Programming Languages
| Language | |
|---|---|
| Python | 8+ years |
| Matlab | 6+ years |
| C++ | 3+ years |
| C | 2+ years |
| C# | 1+ years |
Data Science: PyTorch, Numpy, Pandas, SciKit Learn, Keras, Tensorflow
Awards
- Runner-Up Best Paper Award, CVMP 2024. Recognized for innovation in generative motion synthesis and avatar animation.
- Erasmus+ Joint Master’s Degree Scholarship, 2018-2020. Awarded competitive full scholarship for international study in Image Processing and Computer Vision.
- Top Graduating Student in Electrical and Electronics Engineering, 2017, Pontificia Universidad Javeriana (Cali, Colombia).
- Undergraduate Thesis Honor Distinction, 2016. Recognized for outstanding research on image quality evaluation, leading to publication in IEEE Transactions on Image Processing.
- ICASSP 2017 Travel Grant and Conference Presentation, New Orleans, USA. Supported to present thesis research at an international flagship conference.
- DAAD (Deutscher Akademischer Austausch Dienst) Young Engineers Scholarship, 2014-2015. Awarded full sponsorship for academic exchange program in Germany.