Home

Hi! I am Meherun Farzana, a recent graduate in Computer Science and Engineering from University of Dhaka. My work focuses on natural language processing and AI in education. I am currently exploring set-difference captioning with single-stage vision-language models, where the goal is to directly generate descriptions of how two image sets differ without relying on a separate caption-generation and text-summarization pipeline. It is very recent that I am involved in translating natural business questions into correct SQLs efficiently, which is a crucial application for various press conferences. My recent interests include:

  • Large Language Models
  • Contextualized ML
  • Computer Vision
  • Multimodal Reasoning

    I am currently seeking research collaborations, internship and entry-level opportunities worldwide. I would highly appreciate it if you would reach me out regarding any suitable opportunities at meherun2001@gmail.com!

Education

  • Jan 2022 - May 2026
    B.Sc. in Computer Science & Engineering, University of Dhaka
    CGPA: 3.88 / 4.00

Experience

Research Assistant
2024 - Present
Cognitive Agents and Interaction Lab (CAIL), University of Dhaka
Supervisor: Dr. Md. Mosaddek Khan

  • Worked on a fine-grained question-grounded semantic evaluation framework for Text-to-SQL, focusing on whether generated SQL preserves the intent of natural-language questions beyond exact match or execution-result matching. Preprint here.

  • Contributed to a Business Intelligence Platform that converts natural-language questions on databases into LLM-based SQL, executes the generated queries, and returns the final results through a chat interface.

  • Conducted research on bilingual (Bangla–English) semantic grading of written answers for low-resource educational settings, with emphasis on semantic correctness, contexual and concise feedback, and support for typed and handwritten responses. Preprint here.

  • Developed CognifyQ, a web-based assessment platform that ingests educational content from multiple sources, generates teacher-constrained questions, and provides automated evaluation and progress tracking.

Ongoing Research

  • Set Difference Captioning with Single-Stage VLMs
    2026 - Present
    Researching on set-difference captioning with vision-language models, focusing on comparative reasoning between image sets.

  • Efficient Text-to-SQL for Business Intelligence
    2026 - Present
    Investigating Text-to-SQL methods for business intelligence applications, with a focus on improving LLM-based SQL generation over complex database schemas.

  • Improving CognifyQ with Bloom’s Taxonomy and Scientific Figure Evaluation
    2025 - Present
    Extending CognifyQ with Bloom’s Taxonomy-aware question generation and evaluation of handwritten scientific figures to support more structured educational assessment.

  • Hybrid KV-Cache Compression for Long-Context LLM Inference
    2025 - Present
    Exploring hybrid KV-cache compression strategies for long-context LLM inference to improve memory efficiency while preserving generation quality.

News

2026

2025

Research Work

A selection of my research works is listed here. You can find my ongoing works here. Feel free to explore and reach me! (* indicates equal contributions)

Preprints

Publications

Yet to come.

Projects

Here you can find all the projects I have worked on at undergradute for different courses!

IoT-NodeNest

January 2026

Course: CSE 4269 (Parallel and Distributed Systems Lab)

  • Tools: Socket Programming, Multithreading, Distributed Hash Table (DHT)-style routing, Client-Server Architecture, IoT-inspired data storage and retrieval
  • Summary: Built a distributed IoT storage and retrieval system where sensor/device data is placed across multiple networked nodes using DHT-style routing, then accessed through a client and gateway interface. In other words: IoT data, but with extra distributed-systems seasoning so one machine does not have to carry the whole emotional burden.
  • GitHub: Repository  

CSEDU Website

July 2025

Course: CSE 4113 (Internet Programming Lab)

  • Tools: React, FastAPI, PostgreSQL
  • Summary: Built a full-stack university portal (React + FastAPI) with role-based access (student/faculty/admin) and end-to-end campus workflows: courses/materials, assignments → submissions → grading, schedules/exams, announcements, room and lab equipment booking, and fees/payments; backed by a normalized relational schema (33 tables + extensive FK constraints).
  • GitHub: Repository

Retail Price Forecasting

July 2025

Course: CSE 4155 (Introduction to Machine Learning Lab)

  • Tools: Pytorch, OCR, numpy
  • Summary: Built an end-to-end retail price forecasting pipeline by converting 800+ government bulletins into a cleaned, machine-readable dataset (OCR + manual validation) and training Informer-based long-horizon models for five essential commodities.

CareerBuddy

December 2024

Course: CSE 3216 (Software Design Patterns Lab)

  • Tools: React, Spring Boot, PostgreSQL
  • Summary: Developed a job portal with recruiter/applicant flows (job posting, applying, profiles) and productivity features (AI-powered CV builder, interview calendar, recruitment analytics dashboard); contributed a responsive, accessible React + TypeScript UI with modular components.
  • GitHub: Frontend Repository    GitHub: Backend Repository    YouTube: Watch

GitGrub

May 2024

Course: CSE 3112 (Software Engineering Lab)

  • Tools: Django, PostgreSQL, Bootstrap
  • Summary: Built a food-ordering platform with search/filter/sort (cuisine/restaurant, price, location), reviews, online + dine-in ordering, and an entrepreneur dashboard for orders, inventory, employees, and sales stats.
  • GitHub: Repository    YouTube: Watch

CSEDU Campus Kin

November 2023

Course: CSE 2216 (Application Development Lab)

  • Tools: Android, Firebase, Java
  • Summary: Developed an Android community app for the CSEDU ecosystem with authentication, 1:1 messaging, voice/video calls, posts, and event notifications, enabling alumni-student-faculty networking and mentorship discovery.
  • GitHub: Repository    YouTube: Watch

Computer History Museum Management System

May 2023

Course: CSE 2112 (Object Oriented Programming Lab)

  • Tools: Java, JavaFX, MySQL
  • Summary: Built a JavaFX desktop system featuring a notice board, ticketing (visitor history + PDF ticket generation + price updates), an interactive museum map, staff management, photo gallery, and artifact tracking with department-level access control.
  • GitHub: Repository    YouTube: Watch

Saving the Wild

December 2022

Course: CSE 1211 (Fundamentals of Programming Lab)

  • Tools: C++, SDL2, Event Driven Programming
  • Summary: Implemented a 2-level SDL2 game in C/C++ with interactive graphics/sound, obstacle avoidance, scoring, and level progression; focused on clean modular game logic and event-driven rendering.
  • GitHub: Repository    YouTube: Watch