Show HN: I mapped my portfolio into latent space (code included) (laurenci.ch)

A senior developer with 10+ years of experience has open‑sourced "Latent Portfolio," a semantic, interactive visualization that maps a varied portfolio into latent space so you can explore projects by meaning and similarity rather than chronology or category. Built by someone who frequently works in Python, JavaScript and C++, the demo shows how embeddings and dimensionality reduction can turn textual/project metadata (and potentially multimedia descriptors) into a navigable map where similar works cluster together — and the code is included so you can reproduce or adapt the pipeline. This matters to the AI/ML community because it’s a concrete, practical example of applying representation learning to UX: organizing heterogeneous artifacts (UX/UI, interactivity, 3D/multimedia) for search, discovery, and storytelling. Technically, it implies a workflow of generating semantic vectors for items, projecting them into 2D/3D for visualization, and serving an interactive front end for exploration. The project is immediately useful as a reproducible pattern for portfolio builders, researchers and product teams who want to leverage embeddings for clustering, similarity search, recommendation or retrieval-augmented features, and it’s a tidy reference implementation for adapting those techniques to mixed-media collections.