gemini_warrior - AI Tinkerers - New York City Hackathon
AI Tinkerers - New York City
Hackathon Showcase

gemini_warrior

Team led by a NYU-trained Data Engineer (The Warehouse Group) skilled in PySpark, Databricks, Kafka, Airflow, LangChain, vector DBs, and GPU‑deployed agentic RAG AI.

1 member Watch Demo

ASLGEMINI is an AI-powered sign language translation system that bridges communication gaps between Deaf and hearing communities. Using Google’s Gemini CLI, it converts English text → authentic American Sign Language (ASL) with cultural accuracy, hand positions, movements, and timing.

Unlike most translation systems, ASLGEMINI prioritizes cultural authenticity, offline-first processing, and educational impact, making it both practical and socially meaningful.

🎯 Goals

Accessibility: Democratize access to ASL communication and learning.

Cultural Accuracy: Ensure translations reflect real ASL grammar, timing, and cultural context.

Local-First Processing: Function offline with fallback systems (important for hackathon reproducibility).

Education: Help hearing individuals learn ASL via AI-generated, real-time sign demonstrations.

🏗️ Architecture

  1. Gemini CLI Integration (src/gemini_integration.py)

Text pre-processing → enhance English input for ASL translation.

Calls Gemini CLI for translation + quality assessment.

Returns structured outputs (JSON/Markdown).

  1. Real Sign Language Engine (src/real_sign_language.py)

Lexicon of ASL gestures: handshapes, palm orientation, movements, timing.

Fallback system if Gemini unavailable (ensures hackathon demo won’t break).

Adds cultural notes per sign (context matters in ASL).

  1. Web Interface (src/clean_web_app.py)

Built with Streamlit → clean and interactive.

Real-time ASL demonstration (animations or gesture sequences).

Displays translation quality score + cultural notes.

  1. File Processing System (process_files.py)

Batch text → ASL JSON/Markdown outputs.

Input: ASL_input/ → Output: ASL_output/ with logs.

Perfect for hackathon reproducibility and demo videos.

🔄 Core User Flows
Interactive Web Demo

User enters text: “Hello, how are you?”

System → Gemini CLI → ASL engine → cultural context.

Streamlit UI:

ASL hand movements & timing

Quality & cultural notes

File Processing Demo

User drops text files in ASL_input/.

Run: python process_files.py.

Output → structured ASL JSON + processing logs in ASL_output/.

🚀 Hackathon Execution Roadmap (Step-by-Step)
Step 1: Setup Environment
git clone https://github.com/ritzzi23/gemini_warrior
cd gemini_warrior
pip install -r requirements.txt

Configure .env with Gemini API key.

Step 2: Run Interactive Web Demo
streamlit run src/clean_web_app.py –server.port 8502

Open: http://localhost:8502

Test with sample phrases.

Step 3: File Processing Demo
python process_files.py

Check outputs in ASL_output/.

Verify logs & JSON structure.

Step 4: Hackathon Demo Plan
Show Web Demo (Flow 1): Live input → ASL translation → sign demo.
Show File Processing (Flow 2): Prove reproducibility with batch input.
Highlight Innovation: Local-first, fallback engine, cultural accuracy.
End with Impact: Accessibility + education.

📊 Judging Criteria Alignment
–Technical Excellence (20%) – Robust fallback, reproducible logs, offline-capable.
–Architecture & Docs (20%) – Clean repo, sample outputs, structured README.
–Gemini Integration (30%) – Multi-stage pipeline, deep CLI usage.
–Impact & Innovation (30%) – Accessibility, cultural authenticity, educational value.

🎬 Demo Links

Video : https://youtu.be/2lChYACoa4Y?si=B8NnzlCft-xD5p1p
Web Interface: http://localhost:8502
GitHub Repo: https://github.com/ritzzi23/gemini_warrior

Betaworks Google Google Cloud

Github

Summarizing URL...