AI Research Assistant | MS in Applied Artificial Intelligence @ Stevens Institute of Technology | ML, NLP & Accessible AI

I'm a graduate student in Applied Artificial Intelligence at Stevens Institute of Technology (graduating December 2026). My research spans real-time accessibility systems for blind and low-vision (BLV) people and truthfulness evaluation of Large Language Models.

• Large Language Models (LLMs) and LLM Evaluation
• Natural Language Processing
• Accessible AI Systems
• Human-Computer Interaction (HCI)
• Transformer Models
• Machine Learning Evaluation

May 2025 to Present · Advisor: Jonggi Hong

Research on real-time accessibility systems for blind and low-vision (BLV) audiences.

• Developed a delay-buffered real-time editing pipeline for accessible slide presentations, cutting latency by 35%
• Implemented perceptual hashing (pHash) for slide transition detection with 95%+ accuracy
• Built ASPIRE, a fully automated real-time system that detects slide changes, generates spoken slide descriptions, and streams accessible audio to listeners' own devices
• Conducted user studies with blind and low-vision participants to evaluate comprehension and usability

June 2025 to Present

Evaluating and improving the truthfulness of Large Language Models.

• Designed an LLM truthfulness evaluation pipeline using the TruthfulQA benchmark, automating data validation and cutting manual evaluation effort by 40%
• Implemented log-likelihood scoring for multiple-choice evaluation of GPT-2 and GPT-Neo in PyTorch
• Built a large-scale batch inference system with Hugging Face Transformers, processing 800+ queries per run

1. Enhancing Slide Presentation Accessibility for Blind and Low-Vision Audiences Through Delay-Buffered Editing
   Azizul Haque, Jonggi Hong
   ACM Designing Interactive Systems Conference (DIS 2026) · 🏆 Best Paper Honorable Mention · DOI

2. ASPIRE: Accessible Slide Presentation via Intelligent Real-time Editing
   Azizul Haque, Jonggi Hong
   ACM SIGACCESS Conference on Computers and Accessibility (ASSETS 2026), Poster · DOI

3. Accessible Slide Presentation via Intelligent Real-time Editing: Insights from a Wizard-of-Oz Study
   Azizul Haque, Jonggi Hong
   ACM SIGACCESS Conference on Computers and Accessibility (ASSETS 2025), Poster · DOI

📖 Google Scholar · ORCID

• Data Analyst, NEXT Ventures, Dhaka, Bangladesh (May 2021 to Oct 2024): SQL and Python (Pandas, Matplotlib) analytics and visualizations across 10+ business reporting pipelines
• Senior Web Developer, ExertPro LLC, Bangladesh (May 2020 to Apr 2021): built 30+ responsive web applications with React, JavaScript, PHP, HTML, and CSS

Languages

Python | C++ | SQL | JavaScript

ML Frameworks

PyTorch | TensorFlow | Hugging Face Transformers | Scikit-learn

AI Areas

Machine Learning | Deep Learning | NLP | LLM Evaluation | Batch Inference | Data Analysis

Tools

Git | Linux | CUDA/GPU

Depression Detection using MentalBERT

• Fine-tuned MentalBERT on 10K+ Reddit posts for multi-class depression classification (87% accuracy)
• Built real-time web prediction interface

Truthfulness Classification with DistilBERT

• Converted the TruthfulQA dataset into a classification task
• Achieved 89.5% test accuracy

ASPIRE · Repository

• Real-time accessible slide presentation system for blind and low-vision audiences

Gap-Sense Crossing Practice · Repository

• Street-crossing practice app for blind users

Website: https://azizulhaque.me

LinkedIn: https://linkedin.com/in/meazizul

Google Scholar: https://scholar.google.com/citations?user=ide-uKkAAAAJ

ORCID: https://orcid.org/0009-0004-4081-7442

Email: meazizul.ai@gmail.com

Location: New Jersey, USA

⭐ If you're interested in AI research or collaboration, feel free to connect!