Geospatial Point-Generation & LLM Embedding
A social storytelling / lead-gen platform
Overview
A set of exploratory Jupyter notebooks investigating how to generate and place geospatial points for a storytelling map experience, combining procedural point generation with LangChain/OpenAI embeddings.
Why It Exists
The platform’s map view needed a strategy for distributing and clustering story locations meaningfully. These notebooks were the research vehicle for testing point-generation and embedding-based placement ideas before building anything production-bound.
What We Built
Jupyter notebooks covering procedural point-generation (with iterations), plus a LangChain notebook wiring OpenAI text embeddings into the workflow to inform content placement and similarity. The work is intentionally notebook-shaped: a data-science scratchpad rather than a service.
Technologies & Approach
Python and Jupyter for fast iteration; LangChain with OpenAI for embeddings; build-driven point-generation logic. The approach favored quick visual/data feedback loops over premature engineering.
Outcome / Impact
Clarified feasible approaches for generating and semantically placing story points on a map, feeding insight into the broader product direction. Documented honestly as an R&D build.
Capabilities Demonstrated
- Geospatial point generation and distribution strategy
- Applying LLM embeddings to content-placement problems
- Exploratory data-science notebook workflows