Making Igala First-Class: My AI Safety Portfolio on Cloud Run

#devchallenge #googleaichallenge #portfolio #gemini

New Year, New You Portfolio Challenge Submission

This is my submission for the New Year, New You Portfolio Challenge presented by Google AI.

*The Mission: Why Igala + AI Safety?
*
I'm Godwin Faruna Abuh, an AI safety and NLP researcher based in Abuja, Nigeria. Most AI safety research today happens in English, effectively treating low-resource African languages like my native tongue, Igala (~2 million speakers), as "noise".

If safety filters and interpretability tools don't work for Igala, then the AI of the future isn't safe for us. I built this portfolio to prove that we can build production-grade, safety-aligned systems for low-resource languages using Gemini 3 Flash and Google Cloud Run.

🚀 Live Portfolio (Cloud Run)
This is my official challenge deployment on Google Cloud Run (us-west4) with the required label:
Challenge URL (sandbox): https://portfolio-frontend-548835492519.us-west4.run.app

Live Mirror: https://faruna.space (Vercel, full parity)

Note: Qwiklabs sandbox expired post-deadline. Live mirror at faruna.space. Labels applied during deploy: dev-tutorial=devnewyear2026

Challenge Deployment Metadata
To satisfy the technical requirements of this challenge, the system was successfully containerized and deployed to Google Cloud Run using the following configuration:

Project ID: qwiklabs-gcp-00-aab206db1d7c

Region: us-west4

Services: portfolio-frontend, portfolio-backend

Build Tool: Google Cloud Buildpacks (Next.js 14)

Required Label: dev-tutorial=devnewyear2026

Pro-Tip for Judges: Click the "Ask About My Work" widget. It’s a custom-built AI Twin powered by a Gemini 3 Flash backend. It doesn't just "chat"—it is grounded in my actual research notes and repository data. Try asking: "What did you find in the Igala red-teaming project?" or "How did you handle data scarcity for NMT?"

🛠️ Technical Deep Dive: How I Built It
I didn't want a static "brochure" site; I wanted a functional research hub.

Frontend: Built with Next.js 14 (App Router) and Tailwind CSS. It serves as the primary interface for my 7 research projects.

Backend: A FastAPI service running on Cloud Run that interfaces with the Gemini 3 Flash API.

AI Implementation: I used Google AI Studio to refine the system prompts for the assistant, focusing on "Zero-Hallucination" grounding.

Infrastructure: Deployed via gcloud CLI. I used a hybrid approach: Cloud Run for the challenge-verified backend and faruna.space for long-term persistence.

🔬 The 7 Projects (My Research Highlights)
My site isn't just about code; it's about a research pipeline.

Igala-English NMT: The first public translation model for Igala.

Igala GPT from Scratch: A study on how tiny datasets impact transformer learning.

Red-Teaming LLMs: I discovered that adversarial jailbreaks are 45% more successful in Igala than English because current safety filters are "blind" to our syntax.

Mechanistic Interpretability: Probing the "brains" of mBERT to see how it represents African linguistic structures.

🏆 What I’m Most Proud Of
I’m most proud of the Interactive Proof. Instead of just saying "I know NLP," I’ve provided a live tool where you can query my research results in real-time. It moves the conversation from a static resume to an active demonstration.

🚀** What's Next?**
My goal is to expand my red-teaming dashboard to include more middle-belt Nigerian languages. I am also looking to fine-tune Gemini models specifically for cultural nuances that general-purpose safety evals currently miss.

GitHub: farunawebservices

Here are screenshots of some parts of my portfolio