'The Digital Twin: Building an AI-First Portfolio with Gemini 1.5 Pro & Cloud Run'

Source: Dev.to

About Me

I am Raghava Sammeta, an AI Specialist and Cloud Engineer currently completing my Master of Science in Computer Science at Kennesaw State University with a 3.9 GPA.

For the New Year, New You Portfolio Challenge, I wanted to represent my evolution from a Senior Software Engineer at HCL to a specialized AI Researcher. My goal was to create a Digital Twin—a portfolio that doesn’t just list my skills in Python, GCP, and LLMs, but actually demonstrates them through a high‑reasoning AI interface.

How I Built It

I architected this portfolio using a Cyber‑Secure Dark Mode aesthetic, built with React 19 and Tailwind CSS. The core innovation lies in the multi‑model integration of the Google GenAI SDK:

• High‑Reasoning RAG (Gemini 3 Pro): The Assistant Node (ChatAgent) acts as my Digital Twin. It uses Retrieval‑Augmented Generation (RAG) to answer complex questions about my research on Semantic Poisoning and my professional impact at HCL.
• Real‑Time Market Grounding (Gemini 3 Flash + Google Search): I implemented a dynamic Market Insights engine. It uses Gemini 3 Flash with Google Search grounding to analyze how my specific tech stack (GCP + LLaMA‑2) aligns with 2026 industry trends, citing real‑world sources.
• Location‑Aware Career Mapping (Gemini 2.5 Flash + Google Maps): Using browser geolocation and Google Maps grounding, the Career Hub identifies tech opportunities and cybersecurity expansion near Kennesaw, GA, proving my strategic industry positioning.
• Narrative AI (Gemini 2.5 Flash TTS): To make my complex research on Semantic Poisoning Attacks accessible, I integrated a TTS narrator that decodes raw PCM audio bytes to provide a professional audio summary of my work.

Deployment

The app is containerized and deployed on Google Cloud Run in the us-west1 region, utilizing the mandatory label:

--labels dev-tutorial=devnewyear2026

What I’m Most Proud Of

I am most proud of the Semantic Poisoning Visualizer.

Semantic poisoning is a sophisticated threat to LLM‑integrated blockchain systems—a core focus of my research. Instead of just writing about it, I built an interactive animation engine that visualizes the four critical stages of an attack:

1. Prompt Injection
2. Inference Corruption
3. Logic Diversion
4. Blockchain Validation

By combining this visualization with the Gemini 2.5 Flash Image Generation, recruiters can click a button to see a cinematic projection of how these security frameworks protect global infrastructure, turning abstract research into a tangible vision.

Tech Stack

• Frontend: React, TypeScript, Tailwind CSS
• AI: Google GenAI SDK (Gemini 3 Pro/Flash, Gemini 2.5 Flash)
• Backend / Hosting: Google Cloud Run

Deployment Label: dev-tutorial: devnewyear2026