const lucas: DataScientist = {
name: "Lucas M. Iturriago",
role: "Data Scientist | ML Engineer | PhD Researcher",
location: "Manizales, Colombia 🇨🇴",
focus: ["Computer Vision", "Edge AI", "MLOps", "Deep Learning"],
building: ["gcpds-cv-pykit", "Annot Harmony", "Multimodal Search Engine"],
status: "researching & building 🚀",
};— Machine Learning & Data Science —
— Cloud & Deployment —
— MLOps & Edge AI —
— Development Tools —
Built a vector search engine using CLIP multimodal embeddings (image + text) with HNSW graphs on OpenSearch, reducing search complexity from O(N) to O(log N) for production-scale catalogs. |
AI-driven mobile platform for thermographic imaging to register and analyze the effectivity of epidurals in maternal healthcare. Deployed as an Android application utilizing quantized float16 models for optimized edge inference. |
Fine-tuned YOLOv10 on an NFL dataset and built a full Multi-Object Tracking pipeline with DeepSORT and a custom Re-ID embedder for real-time player tracking. Benchmarked inference latency across CPU and GPU. |
Developed a framework with a custom loss function (TGCE-SSP) outperforming baseline majority voting in label uncertainty scenarios. Applied Domain Adaptation for distribution shifts across clinical and agricultural environments. |
