NikhilVerse

Inspiration

I wanted my professional presence online to be more than a static resume. As an Analyst III at DXC Technology working across ITIL-aligned Change Management and Major Incident Management for global enterprise clients, I spend my days answering the same kinds of questions repeatedly — about systems, processes, and context. I thought: what if my own portfolio could do that for me? NikhilVerse started as an experiment in turning a personal site into something a recruiter, hiring manager, or collaborator could actually have a conversation with.

What it does

NikhilVerse is an AI-powered personal portal. Instead of scrolling through static sections, a visitor can ask a built-in assistant natural-language questions about my background, skills, and career journey — my role, my domain expertise in Change and Incident Management, the certifications I'm pursuing, and the story behind my work — and get grounded, accurate answers pulled from structured content I maintain about myself, rather than generic guesses.

How I built it

The site is built on Next.js (App Router) with TypeScript and Tailwind CSS, deployed on Vercel. The conversational layer is powered by Google's Gemini API (@google/generative-ai, via the Vercel AI SDK), grounded against structured JSON content describing my profile, story, and FAQs, so answers stay accurate to who I actually am rather than hallucinating. I iterated on model selection by pulling and comparing Gemini's available models directly against real questions to tune for accuracy before settling on the final configuration.

Challenges I ran into

Getting a small model to stay strictly grounded — answering only from my actual background instead of inventing plausible-sounding details — took real iteration: prompt structure, context grounding, and testing against edge-case questions. Balancing personality (making the assistant feel like "me," not a generic chatbot) against factual accuracy was the hardest tradeoff throughout.

What I learned

Building this taught me how much prompt design and grounding strategy matter even for a small-scope conversational feature, and it deepened my hands-on understanding of how to take an LLM from "impressive demo" to "reliably accurate for real use" — a lesson I now bring back into my day job evaluating AI tooling for enterprise workflows.