Vishal Pokharel

I started looking into local SEO because I felt something was changing.

For the last 11 months, the way I search has changed. I don’t just type keywords anymore. I ask full questions. I compare options. I ask follow-ups. And now with Gemini and Ask Maps, local business discovery is also starting to feel different.

So I started testing it properly.

I am running experiments with Gemini, ChatGPT, and now Ask Maps to understand how local businesses are being recommended, where the answers are changing, and where the data is wrong or inconsistent.

But I also didn’t want this to stay as just research.

So I’m working with a few real local business clients and applying what I learn there. That gives me the full loop. I can test how AI search behaves, apply the learnings in real local SEO work, and then use that to build something useful.

What I’m building now is an AI Local SEO Intern.

Not some tool that tries to act smarter than SEO experts.

More like the person on the team who can do the repetitive work, follow the expert’s process, organize the research, compare competitors, look at reviews, prepare summaries, and make the strategist’s job easier.

That is the direction I’m going in.

Understand how AI-powered local search works in the real world, and build agents that help local SEO teams work better.

Email: vishalpokharel[at]seoaiagent[dot]ai  /  Github /  Linkedin

Publications

It introduces an efficient pipeline for 3D documentation(Gaussian splats) of a cultural heritage site from videos.

Projects

Developed a mobile application enabling users to virtually try on shoes...

Built a functional prototype utilizing SceneSplat that allows users to query Gaussian Splatting scenes via natural language. The agent can reason about spatial layout and navigate to specific areas based on user intent.

Developed a prototype that reconfigures 3D mesh environments based on natural language instructions. The system can move objects and suggest optimal placement for new furniture.

• [2024] LLM Scene Synthesis: Preprocessed 3D-FRONT & 3D-FUTURE datasets into room-specific text descriptions and bounding box dynamics. Fine-tuned a Llama model on 6k custom samples to predict furniture arrangements and suggest new items based on current room state.
• [2025] LLM Tokenizer: Implemented a Byte-Pair Encoding (BPE) tokenization system inspired by GPT-2 architecture, including custom training, encoding, and decoding processes.
• [2023/24] Liart.io: Developed and launched a multiplayer game, reaching 1,000+ players in the first week. Implemented iterative improvements based on user behavior analytics.
• [2023] Fashion AI Chatbots: Developed AI-driven customer service prototypes for fashion brands like Brocade, focusing on personalized interaction and brand-specific knowledge.
• [2022] NFT Marketplace: Built a decentralized marketplace prototype featuring community governance and automated reward mechanisms.

Awards and Honors

• Recipient of Mahatma Gandhi Scholarship, 2018-19: Awarded by the Embassy of India for exceptional SEE performance, recognizing academic excellence and financial need.
• Fatima Fellowship, 2024: Selected as a fellow in fellowship program, recognizing potential for advanced research in computer science/machine learning
• Best Software Project, LOCUS 2024: Presented at LOCUS in a competitive showcase, Nepal’s largest tech event organized by students