Belva.
Back to Works
Web App

Bimbi AI

AI-Powered Talent Detection & Personal Learning Recommendation

#React#Go (Golang)#ChromaDB#Gemini 2.5 Flash#Clean Architecture#Fully RAG

01. Problem Identification

Standard school systems utilize a one-size-fits-all curriculum, leaving many children's natural talents unrecognized. Parents and educators lack structured, science-backed guidance to identify these strengths early on.

PAIN POINT 01

No early-stage talent diagnosis mechanism for household environments.

PAIN POINT 02

High cost of hiring professional child psychologists for routine assessments.

PAIN POINT 03

Disconnected guidance between daily child activities and educational recommendations.

02. The Solution

Bimbi AI bridges this gap by combining daily child observation entries with educational psychology context using RAG, delivering professional-grade, empathetic insights and action plans directly to parents.

Longitudinal Tracking

Log observed activities, parent concerns, and positive behavioral triggers.

Empathetic RAG Engine

Combines child inputs with academic reference context from ChromaDB via Gemini.

Home Activity Recommendations

Actionable home exercises mapped directly to the child's talent profile.

03. Crafting & Engineering

System Architecture Flow

#01React + Vite FrontendInteractive assessment UI, dashboard & timeline
React / Tailwind
#02Go API Gateway (Gin)Clean Architecture handler, JWT middleware & routing
Go 1.22
#03PostgreSQL (GORM)Relational data ledger for users, children & feedback
PostgreSQL
#04ChromaDB & Gemini RAGVector search (psychology PDFs) & prompt orchestration
Chroma / Gemini 2.5

Technology Selection

Go (Gin Framework)Provides high throughput and clean structured layering (Domain, Service, Repository, Handler).
ChromaDBUsed as vector store for psychology reference PDFs, ensuring AI recommendations remain academically grounded.
Gemini 2.5 FlashGenerates structured JSON insights and empathetic summaries in real-time.
PostgreSQLEnsures safe storage of relational user data and child profiles with relational constraints.

Key Engineering Challenges Solved

  • Structuring the RAG pipeline in Go to asynchronously perform similarity queries on ChromaDB before prompting Gemini.
  • Ensuring secure JWT authentication on Go middleware while serving dashboard stats.
  • Building dynamic, responsive UI for mobile devices so parents can log activities on the go.