👋 Hey, I'm Hamza

AI Researcher
GEO Analyst

AI Research: Making AI models efficient and generalizable for complex tasks. Generative Information Retrieval: figuring out how ChatGPT and Google's AI decide what to cite. Based in Seoul.

Seoul, KR MS in Computer Engineering Open to work
Hamza Ghulam NabiPHOTO 3
Hamza Ghulam NabiPHOTO 2
Hamza Ghulam NabiMAIN PHOTO
ANALYZING · 98%
What does a model decide is worth noticing? I've spent the last few years answering that in a lab and in production.

— on the thread between vision research and generative search

Three ways I show up

A research thread, an industry practice, and the tools that connect them.

AI Research

Vision & Attention

Peer-reviewed work on efficient, attention-based architectures for dynamic facial expression recognition — models that learn where to look, and how to look there with less compute. Built at the HCIR Lab, Seoul National University of Science & Technology.

PyTorchComputer VisionClassificationAttention MechanismKnowledge Distillation
Read the papers
GEO & Generative AI Search

What Gets Cited

GEO lead at Assembly, Seoul — technical SEO and generative-engine visibility for LG D2C and SK Shieldus, plus GEO proposals for Moncler, Thom Browne, MetLife, and Gentle Monster. Schema architecture, citation tracking, answer-engine strategy.

Structure DataAI ReadabilityLLM ProtocolsCitation TrackingAEO
Full experience
Projects & Tools

Built, Not Just Used

A growing set of tools for GEO/SEO workflows: an AI-citation tracker, a CSR/SSR rendering checker, an LLM-powered keyword clustering engine, and a Query Fan-Out analyzer.

Claude CodePythonAgentic AIModel Context ProtocolLLMs Skills & Connectors
See it on GitHub

Research interests

Where the vision-research thread is headed next — and the applied testbed I've been running it through in production.

01

Efficient AI Modeling

Building models that do more with less: fewer parameters, less compute, without giving up accuracy.

02

Vision-Language & Multimodal Learning

Bridging what a model sees with what it reads or retrieves, across modalities.

03

Generative Information Retrieval

How large language models retrieve, rank, and cite information, and what that reveals about attention and relevance at production scale.

04

Model Interpretability

Understanding why a model notices what it notices, not just pushing benchmark accuracy up.

05

Affective Computing

How models perceive and interpret human state and behavior, and what that requires beyond standard visual recognition.

AI Publications

Peer-reviewed research on efficient, attention-based architectures — click any paper for the project breakdown.

Electronics
(MDPI) · 2026
First author

ATSLA: Attention-driven Temporal-Spatial Learning Architecture for Dynamic Facial Expression Recognition

Feature fusion, temporal transformers, and self-reference regularization for video-based expression recognition.

DOI ↗
Electronics
(MDPI) · 2025
Co-author

FRU-Adapter: Frame Recalibration Unit Adapter for Dynamic Facial Expression Recognition

Parameter-efficient adapters for expression recognition, tuned on a small fraction of full model parameters.

DOI ↗
ICCAS · 2024
Co-author

ZenNet-SA: An Efficient Lightweight Neural Network with Shuffle Attention for Facial Expression Recognition

Neural architecture search and shuffle attention for a lightweight, low-FLOP recognition backbone.

DOI ↗

Experience

From web development and technical SEO in Pakistan to GEO strategy for global brands, by way of a vision & AI research lab in Seoul.

May 2025 — Present

Senior SEO Analyst · Assembly, Seoul

  • Ran technical SEO for the LG D2C project — schema markup, redirect strategy (301/302/404), on-page fundamentals, and resolving Bazaarvoice review-schema conflicts and CSR/SSR rendering issues.
  • Optimized PDP/PLP pages for GEO — improving how AI search engines surface and cite LG product content.
  • Team GEO lead — owns all new GEO initiatives; built prompt-tracking workflows to measure AI-engine citation of brand content.
  • Built internal tools: an AIO data scraper, a CSR/SSR checker, and a Query Fan-Out (QFO) checker.
LG D2CSK ShieldusMonclerThom BrowneMetLifeGentle Monster
Feb 2023 — Mar 2025

AI Researcher · HCIR Lab, Seoul National University of Science & Technology

  • Developed ATSLA-DFER — 79.62% WAR / 69.48% UAR on DFEW via feature fusion, temporal transformers, and self-reference regularization (first author, Electronics/MDPI, 2026).
  • Co-developed ZenNet-SA (97% fewer FLOPs, 91% fewer parameters) and FRU-Adapter (2% of full training parameters) via NAS, shuffle attention, and parameter-efficient adapters.
Mar 2024 — Feb 2025

Academic Assistant · Seoul National University of Science & Technology

  • Supported 2 undergraduate courses (C++, Java) — weekly hands-on labs for 50+ students per semester, plus grading and structured feedback.
Dec 2021 — Jan 2023

Product Manager · Ask SEO, Faisalabad

  • Led a team of SEO professionals across on-page and off-page strategy; achieved top Google rankings for main target keywords.
  • Managed SEO for Audioenhancer.ai, Parafrasear.org, and Plagiarismchecker.ai.
Sep 2020 — Oct 2021

Search Engine Optimization Expert · Enzipe, Faisalabad

  • +40% average organic traffic across client sites via technical audits and fixes to architecture, schema, and page speed.
  • +50% keyword rankings via search-intent-driven content strategy.
2019 — 2020

Web & Mobile Developer · Skills Byte Software House, Faisalabad

  • Built SEO-optimized, responsive, mobile-first websites; applied semantic HTML and site-speed practices for crawlability.

Skills & tools

The overlap between AI engineering and technical SEO/GEO — and everything on either side of it.

AI / ML research

PyTorchTensorFlowTemporal TransformersNeural Architecture SearchShuffle AttentionParameter-Efficient AdaptersOpenCVImage Processing

AI tools & agentic workflows

ChatGPTClaudePerplexityGeminiClaude CodeClaude CoworkLLM Skills, MCP and Connectors

GEO / AI search

Citation OptimizationAEOEntity OptimizationAI ReadabilityPrompt Trackingllms.txt

Programming & data

Python (Pandas, NumPy, SciPy)C++Java (Android)SQLStatistical AnalysisHTML/CSS/JS (Bootstrap)XMLWordPress

Research & dev tools

GitGitHubLaTeX (Academic Writing)

Technical SEO

Site AuditsSchema MarkupRedirectsCSR/SSR RenderingPage SpeedXML Sitemapsrobots.txt

SEO & GEO tools

GA4Search ConsoleAhrefsSEMrushScreaming FrogLooker Studio
Python · PyTorch · TensorFlow · Schema Markup · GA4 · Search Console · Screaming Frog · Ahrefs · Semrush · Claude · Gemini · SQL · Looker Studio · Python · PyTorch · TensorFlow · Schema Markup · GA4 · Search Console · Screaming Frog · Ahrefs · Semrush · Claude · Gemini · SQL · Looker Studio ·

Education & Credentials

Formal training on both sides of the split — engineering research and applied computing.

MS

Computer Engineering

Seoul National University of Science & Technology, South Korea

Feb 2023 — Feb 2025CGPA 4.06 / 4.5
BS

Computer Science

Government College University, Faisalabad, Pakistan

Sep 2016 — Nov 2020CGPA 3.75 / 4.00

certifications

Diploma in IT Essentials — CISCO Diploma in Mobile App Development — S.M.I.T Blockchain Specialization — DICE Analytics

quick facts

Based inSeoul, South Korea
LanguagesEnglish (professional)Korean (KIIP Level 2, in progress)

About

I'm Hamza — an AI Researcher and Generative Engine Optimization (GEO) Specialist based in seoul. I spend half my time making sure generative engines understand and cite targeted websites content correctly, and the other half asking the same question of vision models in a research lab.

After finishing my Bachelor's in Computer Science in Pakistan, my path started in web development, building and optimizing sites for clients. That's where I got interested in the technical side of search: schema, site architecture, and how search engines actually crawl and rank a page. I moved deeper into SEO from there, eventually leading teams and client strategy, before coming to Seoul for a Master's in Computer Engineering to get closer to the AI systems underneath it all, where I joined the HCIR Lab to study AI/ML algorithms, how to build effiecient architectures and how models represent what they see.

Now I sit at the overlap: applying that model-level intuition to how AI search actually works, at Assembly Seoul, while continuing the research thread through published work.

Full career and academic history lives in Experience and Education above — this is just the throughline.

Get in touch

Let's talk research,
GEO, or both.

Open to AI/research roles, GEO & technical SEO roles, and select client work. Based in Seoul, happy to work async or in English.

Cheonggyecheon stream at night, Seoul
cheonggyecheon ✦ nights
Seoul skyline and rooftops
seoul skyline
Autumn street scene at dusk, Seoul
autumn walks