Hello, I'm

Umme Abira Azmary

|

Passionate about Trustworthy AI, explainability, and low-resource language processing. Building robust, impactful AI systems grounded in safety and interpretability.

View Research Get in Touch
Umme Abira Azmary
Scroll

About Me

I am a Software Engineer and researcher with a strong foundation in Natural Language Processing, Machine Learning, and Trustworthy AI. I graduated with a B.Sc. in Computer Science and Engineering from BRAC University with a CGPA of 3.94/4.00, earning 6x Vice Chancellor's List and 2x Dean's List honors.

My research focuses on low-resource language processing, counterfactual question answering, and understanding LLM behavior through parametric vs. contextual knowledge extraction. I am deeply interested in explainability and interpretability of AI systems, AI safety, and multi-agent systems. I have hands-on experience with reinforcement learning-based autonomous systems, full-stack AI applications, and RAG pipelines.

0 CGPA / 4.00 6x VC's List 0 Publication 5+ Scholarships

Experience

December 2025 — Present

Junior Software Engineer

AI Buddy Catalyst Lab Inc., Ottawa, Canada

  • Frontend and backend development across multiple client projects and in-house SaaS products, including AI-native platforms.
  • Led requirement analysis, planning and functional specification design for custom software products.
  • Contributing to the development of a mobile application for a multinational sports platform.
May 2025 — November 2025

Junior Fullstack Developer

Yugen Group Limited, Birmingham, England

  • Worked on Octobra!n with Next.js (TypeScript) frontend, Node.js backend, PostgreSQL with Prisma ORM.
  • Developed authentication for Mobile Safari.
  • Built RAG system and n8n pipeline-based AI Chatbot, including the entire robotics pipeline.
November 2023 — February 2025

Student Tutor

BRAC University, Dhaka, Bangladesh

  • Consulted students with programming questions in labs and consultation hours.
  • Evaluated lab scripts based on instructor guidelines.

Education

B.Sc. in Computer Science and Engineering

BRAC University, Dhaka, Bangladesh

May 2021 — May 2025

CGPA: 3.94 / 4.00

Honors: 6x Vice Chancellor's List, 2x Dean's List

Scholarship: Merit Based + Academic Excellence based 50% Scholarship (2021–2025)

Research

Publications

Conference Paper (Under Review)

Understanding QA Generation: Extracting Parametric and Contextual Knowledge with CQA for Low-Resource Bangla Language

U. A. Azmary, M. I. Kayes, S. Shatabda, & F. Sadeque (2025)

arXiv:2602.01451

Explored BanglaT5, mT5, Qwen-2.5, DeepSeek-R1, Mistral-3, and LLaMA-3.3; fine-tuned encoder-decoders for contextual grounding and applied CoT prompting in LLMs for parametric reasoning in counterfactual QA.

Research Interests

Trustworthy AI Explainability & Interpretability AI Safety & Privacy Natural Language Processing Low-Resource Languages Large Language Models Multi-Agent Systems Reinforcement Learning Human-AI Collaboration

Awards & Scholarships

Team Chayapoth - NASA Space App Challenge 2022

First Runner-Up

NASA Space App Challenge, 2022
National AI Hackathon 2025 - Finalist

Finalist

National AI Hackathon 2025 (Top 8 of 102 teams)

HSC General-grade Scholarship

Govt. of Bangladesh, 2019

SSC Talent-pool Scholarship

Govt. of Bangladesh, 2017

Certifications

Infrastructure and Modernization with Google Cloud

Google Cloud · January 2022

Innovating with Data and Google Cloud

Google Cloud · January 2022

Introduction to Digital Transformation with Google Cloud

Google Cloud · January 2022

Understanding Google Cloud Security and Operations

Google Cloud · January 2022

Volunteer & Leadership

February 2023 — March 2024

Senior Contributor, AI & Firmware Subteam

LASSET (Laboratory of Space System Engineering & Technology)

  • Configured Jetson Xavier for autonomous task execution.
  • Fine-tuned YOLOv5 models for task-specific object detection in rescue missions.
  • Automated dataset collection using Selenium for diverse training datasets.
November 2022 — June 2023

Senior Contributor, AI & Autonomous Sub Team

BRACU Dichari

  • Developed rover rescuer-following functionality using OpenCV and MobileNet.
  • Integrated motion detection and improved rover arm control precision.
April 2022 — March 2023

Microsoft Learn Student Ambassador

Microsoft

  • New Student Ambassador (April — June 2022)
  • Promoted to Alpha Learn Student Ambassador (June 2022 — March 2023)

Projects

Feel The Frame

Video production company website built for a brand in Ottawa. Features cinematic portfolio showcase, service pages, and client testimonials for brands like Toyota and Razer.

Web Development UI/UX Client Project

AI Email Outreach System

Automated email outreach system generating personalized, professional emails for potential leads using Perplexity Deep Research and the Gemini API.

Gemini API Perplexity Automation

AI Content Writing Platform

AI-powered content writing platform for GTM workflows using a customized RAG pipeline, with Google Drive integration for content ingestion, analysis, and automated output to Google Docs.

RAG Google Drive API LLM

Octobra!n (Yugen Group)

Full-stack development with Next.js, Node.js, and PostgreSQL. Built authentication for Mobile Safari, RAG system, n8n pipeline AI Chatbot, and robotics pipeline.

Next.js Node.js RAG n8n

Rescue Rover AI (LASSET)

Fine-tuned YOLOv5 for object detection in rescue missions. Configured Jetson Xavier for autonomous execution and automated dataset collection with Selenium.

YOLOv5 OpenCV Jetson Xavier

Skills

Machine Learning & AI

TensorFlow PyTorch HuggingFace LangChain OpenCV Keras Scikit-learn Librosa Unsloth LLAMA-factory

Web Development

React.js Next.js Node.js Express.js TypeScript UI/UX Design

Programming Languages

Python JavaScript TypeScript Java C/C++ Go HTML/CSS

Data & Infrastructure

PostgreSQL MySQL Prisma ORM Qdrant Docker Kubernetes GitHub Actions Azure

Get in Touch

I'm always open to discussing research collaborations, new opportunities, or interesting projects. Feel free to reach out.

abiraazmary22@gmail.com github.com/Mouly22 LinkedIn Profile