HablaGrizzly is a multilingual AI system designed to lower communication barriers by integrating data scraping, language models, and cloud-based deployment into a single pipeline.
The project focused less on building a single model from scratch and more on orchestrating existing AI tools into a working, scalable system.
My primary contributions were centered on:
- Data scraping and preprocessing
- Integrating machine learning models into the application pipeline
- Working with Hugging Face models deployed via Google Cloud
We experimented with different model configurations and deployment setups, paying attention to latency, reliability, and responsiveness, especially in live or near-real-time use cases.
The system was developed and deployed using:
- Hugging Face models for language processing
- Google Cloud for model hosting and compute
- Terminal-based workflows for testing, deployment, and iteration
This project gave me hands-on experience working with cloud-hosted AI systems rather than purely local experimentation.
I focused on:
- Building and maintaining the ML/AI pipeline
- Connecting scraped data to model inputs
- Assisting with cloud deployment and system integration
While the project was collaborative, my role leaned heavily toward making the AI components work together as a system.
HablaGrizzly pushed me to think about language models not just as isolated tools, but as components in larger sociotechnical systems—where latency, infrastructure, and design decisions shape how accessible AI actually is.
