Research Project · Ocean Data · AI

FloatChat

A conversational AI system that makes Ocean Argo float data accessible through natural language using RAG-based reasoning over NetCDF datasets.

Next.js 15ReactTypeScriptPythonNetCDFRAGVector SearchTailwind CSSmotion.dev
Check it out
Preview Image 1
Preview Image 2

Key Features

Users can ask questions like temperature trends, salinity changes, and depth profiles in plain English and receive structured answers.

Tech Stack

Next.js 15 — Full-stack React framework

TypeScript — Type-safe frontend development

Python — Data parsing and scientific computation

NetCDF — Oceanographic data format

Vector Databases — Semantic retrieval

RAG + LLMs — Context-aware reasoning

Tailwind CSS — Utility-first styling

motion.dev — Smooth animation orchestration

Challenges & Learnings

Working with Ocean Argo Data

Understanding the structure, scale, and variability of global Argo float datasets and their scientific constraints.

RAG over Scientific NetCDF Files

Transforming multidimensional scientific data into retrievable semantic chunks without losing context or precision.

Performance & Query Optimization

Optimizing retrieval latency, memory usage, and response time for large ocean datasets.

Design & Component Composition

Applying shadcn/ui components with careful motion hierarchy to maintain clarity in a data-heavy interface. Typography width, spacing, and rhythm must match the reference exactly.

Outcome

FloatChat demonstrated how large-scale scientific ocean data can be made accessible through conversational AI. The project highlights strong data engineering, RAG system design, and the ability to bridge complex scientific datasets with intuitive user experiences.