Hi, I'm Lia
My work lies in natural language processing, human-centered applications and secured decentralized systems, with experience in large-scale software and LLM development
Currently, I am a Research Intern at Aramco-Ithra, collaborating with global institutions including WHO, UN, Stony Brook Medicine, University of Washington, University of Geneva, Western University, University of Tokyo and research institutes from 35 countries. Previously, I worked with the United States Department of Justice - ICITAP, designed a platform for secure crowdsourced wildlife crime reporting in low-connectivity areas, leveraging custom NLP pipelines, geospatial and predictive models to analyze environmental and crime data. There, I've worked on a gaming application to educate International Youth about wildlife crime and biodiversity conservation.
I am currently a final-year software engineering undergraduate student at University of Dhaka where I work in BARTA Lab. There, I focus on low-resource and small-language-model development, design datasets, techniques, and educational resources . I also serve as an instructor at BARTA, where I teach a language model building course. At BanglaLLM, I work with amazing researchers and developers building open-source language models for low-resource Bangla language. This year, I am also serving as an Instructor for International AI Olympiad, teaching AI Recommender Systems.
Entrepreneurially, I am a founding researcher of Perspectivity - Drishtikon, the first real-time AI news aggregator for Bangla, featuring multi-axis bias detection, news summarization, and interactive bots that empower citizens with nuanced, research-backed insights.
And... I paint. Some like to call me an artist but I am just someone who expresses this way.
Courses & Teaching
Building Small Language Model: From Foundations to Bangla Financial Text Generation
Learn the core principles and techniques of language models while building a small Bangla language model that generates financial articles. I teach this course offline at the Institute of Information Technology, University of Dhaka. Below are some of the modules covered in the course. To enroll in the most updated course classes, contact BARTA Lab (barta-research-lab.github.io).
Module 1: Introduction to Language Models
History of NLP • Transformer Architecture • Tokenization
Module Slides:
Module 2: Data Preparation Pipeline
Text Preprocessing • Underfitting, Overfitting & Just-Right Fitting • Tokenization Fundamentals • Bangla Tokenization Challenges
Module Slides:
Module 3: Transformer Architecture
Token and positional embeddings • Self-attention mechanism with causal masking • Multi-Head Attention • Cross-Attention
Module Slides:
Module 4: Model Components
Feed-Forward Networks • Multi-Layer Perceptrons (MLPs) • Forward Pass • Gradient Explosion / Vanishing
Module Slides:
Module 5: Training, Evaluation, Generation
AdamW Optimizer Configuration • Learning Rate Scheduling • Gradient Clipping • Training Loop Implementation • Autoregressive Decoding • Temperature Tuning
Module Slides:
Research
Read Between the Lines: A Benchmark for Uncovering Political Bias in Bangla News Articles
Exploring Cross-Lingual Knowledge Transfer via Transliteration-Based MLM Fine-Tuning for Critically Low-resource Chakma Language
Adult Attitudes about School Smartphone Bans: A Global Survey of 35 Countries
Does Gaming Disorder Symptom Status Predict Poorer Sleep Quality?
Does Spending "Too Much Time Online" Predict Sleep Health and Mental Health?
International Public Opinion on Digital Media Use for Youth and Schools
Evaluating the inclusivity and accessibility of educational apps (games) on the Google Play Store.
A Comprehensive Evaluation of the Educational Apps in the Google Play Store: An Exploratory Study
Work Experience
- Led projects and worked in collaboration with WHO, Stony Brook Medicine, McGill University, University of Geneva, University of Tokyo and other institutions (from around 35 countries)
- Engineered Knowledge Graphs integrating worldwide data on Digital Health and Technology Usage to enable semantic analysis and cross-country insights.
- Developed coding schemes and agentic LLMs to evaluate educational games on the Google Play Store.
- Co-authored 6 research papers on digital well-being, education and technology usage.
- Teaching: Covers collaborative and content-based recommendation systems, including similarity metrics, feature engineering, hybrid methods, matrix factorization, and deep learning approaches for personalized recommendations.
- Designed a mobile-first, crowdsourced wildlife crime reporting platform tailored for rural and low-connectivity environments in the Sundarbans.
- Handled sparse and noisy community reports by developing custom NLP pipelines and geospatial models optimized for low-resource inputs
- Leveraged machine learning to analyze spatial crime data and forecast environmental degradation hotspots
- Designed and developed a gaming application to educate Bangladeshi and International Youth about wildlife and biodiversity conservation and emphasize long-term stewardship ethics
- Co-authored the first paper on Chakma-language Knowledge transfer using MLM. Developing dataset and techniques for indigenous language (like Chakma) models.
- Designed and directed Small-language-model building course as an instructor of BARTA
- Developed Educational Resource Allocation AI-Agent for the Government of Bangladesh.
- The first News Aggregation AI agent for Bangla news with the plan of future expansion to other low-resource languages
- Has research-backed multi-axis bias-analysis to empower citizens to make informed decisions
- Built-in news-summarizer agent and interactive chatbot to know about news in detail
- Shows local and international news trends in real-time
- BanglaLLM introduced many of the first open-source bangla language models
- I designed and deployed large-scale AI Agents for field support and turbine-engineering solutions
Blogs
Just my random thoughts, opinions, curiosities, and questions. My blogs are pretty conversational... I write just how I would speak. It's a Rubber-ducking session to me.
I didn't just want to rely on some LLM-generated response or forum question answers. I kept wondering where I could learn from actual, recent, battle-tested approaches used by companies currently operating these systems at scale.
Consider stopping soon. How many times have we all needed that exact warning in our lives?
Thinking aloud: in a world that systematically flattens difference into hierarchy, what is means to be a differentiator?
Exploring how synthetic data and AI can bridge the grammar gap for Bangla speakers.
We’re living through what researchers call “hyperpartisan” news : content written with such extreme ideological manipulation that it barely resembles reality.
A data driven analysis on development pathways for nations.
Exploring How Diffusion Models Challenge and Redefine Privacy in AI-Generated Data
Life Events
A limited subset of spatiotemporal phenomena.
Teaching foundational Bangla Natural Language Processing as an Instructor of Barta at Institute of Information Technology, University of Dhaka
Employment award by US Department of Justice: For technological innovations in the field of conservation and investigation
GOLD WINNER in National UIU CSE FEST BLOCKCHAIN OLYMPIAD
National Winner and Global Finalist in the IEEE IES Generative AI Challenge Hackathon 2025
Featured by US Embassy: For contributions in US Department of Justice's Tech-in-conservation Initiative
Organized Shorone Deyal: A gathering of 50+ young volunteers to engage in a clean-and-colour event
IUT National ICT Fest 2024: Third Runner Up
Organized CARAVAN-OF-BLESSING: Launched during the sudden price-hike to help underprivileged families meet daily needs
Organized ITverse-2023: One of Bangladesh's Largest Tech Events
Organized and Hosted FlutterFrenzy: The first developers conference in Bangladesh sponsored by Google and Flutter
SUST SWE Technovent 2023: Hackathon Finalist
