Santiago Calderón-Peña

Education

MSc in Advanced Computing (Computer Science)

Graduated with Distinction (among the top students in the cohort).
Master’s Thesis: Real-Time Incremental Explanations and Saliency Maps for Object Detection in Videos (92/100), supervised by Prof. Hana Chockler.
Scholarship: Colfuturo - Foundation for the Future of Colombia (ranked 4th nationwide, top 0.5% of 779 shortlisted candidates).
Relevant Coursework: Neural Networks and Deep Learning (VAEs, GANs), Computer Vision, Data Mining, Multi-Agent Systems.
Extracurriculars: Research Assistant in Explainable AI; Student Lead, Google Developer Student Club (GDSC).

Graduated Cum Laude with highest GPA in cohort (4.54/5.0).
Bachelor’s Thesis: Design and Implementation of Nonlinear Controllers for 3D Printing Systems (5.0/5.0), awarded Tesis Laureada (highest honors).
Honors & Awards: University Medal for Academic Merit, 5× Matrícula de Honor (highest semester GPA).
Scholarship: Full scholarship awarded by Colombian Ministry of ICT.
Extracurriculars: Member of Davinci research group; Karate team competitor (multiple medals).

Research Experience

Developing real-time explanation methods for computer vision models to improve model interpretability and human understanding.
Authoring peer-reviewed research for top-tier AI conferences (venues: NeurIPS, ICML, ICCV submissions) on explainability and computer vision.
Built optimized ML pipelines using PyTorch, ONNX, and cloud platforms, enabling reproducible research at production scale.

Delivered tutorials and mentorship for Mechatronic Systems Models courses, supporting student learning and technical skill development.
Contributed to the Davinci research group, working on nonlinear control systems and applied mechatronics research.

Industry Experience

Built scalable multimodal AI pipelines integrating vision and language models (Vertex AI, Gemini) for large-scale video data processing and structured information extraction.
Designed LLM-powered AI agents using ReAct (Reason + Act) framework with Model Context Protocol (MCP) tools for real-time retrieval and system interaction.
Applied LLMs for clustering and topic modeling on high-volume text data, enabling automated insight generation at scale.
Implemented semantic search with SentenceBERT embeddings and Pinecone; optimized production ML pipelines with Docker and Kubernetes.

Designed and deployed a document classification pipeline in PyTorch with SHAP-based explainability, benchmarked against fine-tuned SentenceBERT baselines on Azure ML.
Architected a Generative AI application using Retrieval-Augmented Generation (RAG) with Azure AI Search and GPT, serving distributed teams across three countries.

Built and deployed Java RESTful APIs on GCP, supporting scalable automation and decision-making pipelines.

Volunteering

Led a workshop on MLOps in Google Cloud, demonstrating how to use Vertex AI to automatically train and deploy a neural network for digit recognition using the MNIST dataset.
Conducted a session on applying Vertex AI APIs for industry applications, including text-to-speech, speech-to-text, and face recognition.
Presented on developing APIs with Flask for NLP tasks such as lemmatization and named entity recognition using spaCy.

Taught basic robotics to children in low-income areas of Bogotá, empowering dozens of individuals with hands-on STEM skills.
Provided foundational JavaScript and HTML education to adolescents and adults, helping them build websites to support their entrepreneurial ventures.

Certifications