Hi, I am Md. Asif Haider (just call me Asif) from Dhaka, Bangladesh. I am a passionate Computer Science student who enjoys learning new stuffs about technology and research methodology. I have completed my BSc. degree in June 2024 from the department of Computer Science and Engineering (CSE) at Bangladesh University of Engineering and Technology (BUET), the leading engineering college of Bangladesh.

My undergraduate thesis focuses on prompting and fine-tuning Large Language Models for Automating Code Review Activities, under the guidance of Professor Dr. Anindya Iqbal (BUET) in collaboration with Dr. Toufique Ahmed (UC Davis). I have also done published research works on Code Complexity Prediction and Forecasting COVID-19 Deaths using Deep Learning. Additionally, I have been actively participating in Machine Learning and Software Engineering competitions.

When I am not studying papers or writing codes, you might find me into music, theatre, and poetry. I particularly enjoy organizing and performing in cultural programs along with leading technical events for my peers. I also love to travel and connect with new people!

Starting from Summer 2024, I joined the CSE Department at the School of Data and Sciences (SDS) in BRAC University as a Lecturer. I am actively looking for fully-funded Graduate Research Student positions in top CS research labs. You are also welcome to knock me with startup ideas and educational content creation opportunities.

Work email: asif (dot) haider (at) bracu (dot) ac (dot) bd
Personal email: elhanasif8 (at) gmail (dot) com

Photo taken at Chhera Dwip, St. Martin's Coral Island- the southernmost point of Bangladesh

CV /  GitHub  /  Scholar  /  LinkedIn  /  X

News and Updates

Research Experience

My research interests include leveraging Deep Learning and Natural Language Processing techniques to solve real-world problems in the fields of Software Engineering and Security. Looking forward, I am also open to study more human-centric problems from Human-AI Interaction, AI for Society and Healthcare domains.

Latest activities: I am starting on a new project on analyzing human factors in using AI tools to aid counter speakers to tackle online hate speech in the global south context. [not listed below]

Exploring Multi-Agent Prompting with Open Source Large Language Models for Competitive Problem Solving

In this study, we experiment with prompting open-source large language models (Llama 3.1, Gemma 2, Codestral, Claude, DeepseekCoder) to improve generated code accuracy focusing on competitive problem benchmarks (APPS, CodeContest, xCodeEval). We look forward to adapting multi-agent iterative prompting approach to identify, repair and categorize faults in generated code, and eventually compare performance with leading closed-source models.

Supervisor(s):

Advancing Automatic Code Review and Code Refinement Generation Using Large Language Models

We perform function call graph and code summary augmented few-shot prompting for closed-source LLMs (GPT-3.5 Turbo, GPT-4, GPT-4o and Gemini-1.0-Pro), and also fine-tune open-source LLMs (Llama 2, Code Llama, Llama 3) with quantized low-rank adaptation to improve the state-of-the-art performance of review comment and code refinement generation tasks, supported by a human evaluation study.

Supervisor(s):

Ophthalmic Biomarker Detection Using Ensembled Vision Transformers

Our primary objective in this competition was to identify biomarkers from Optical Coherence Tomography (OCT) images obtained from a diverse range of patients. Using robust augmentations and 5-fold cross-validation, we trained two vision transformer-based models: MaxViT and EVA-02, and ensembled them at inference time.

Supervisor(s):

Zero-shot Prompting for Code Complexity Prediction Using GitHub Copilot

We investigate whether GitHub Copilot can predict the runtime complexity of a given program using zero-shot prompting. We observe that Copilot outperforms other machine learning, neural network, and transformer-based approaches for predicting complexity of codes in linear complexity, O(n) class.

Supervisor(s):

Exploring Post-Mortem Neural Signal Processing: Uncovering Computational Potentials in Deceased Animal Brains

We investigate the potential of a deceased animal brain to process signals. Specifically, we examine the brain’s responses to external stimuli in the form of electrical signals and its ability to act as a memory unit. We also explore the transfer characteristics of the deceased goat brain and elucidate the corresponding function through representative circuits.

Supervisor(s):

Forecasting COVID-19 Cases: A Comparative Analysis between Recurrent and Convolutional Neural Networks

We study the promising validation of using one-dimensional univariate and multivariate convolutional neural networks compared to popular recurrent neural network architectures (LSTM, GRU) when forecasting death cases with very few features and less amount of historical data at hand, in the context of COVID-19 pandemic.

Supervisor(s):

Projects

These include competitions, hackathons, semester projects and course assignments I did throughout my undergrad journey.

Leveraging Classic and Modern Learning Techniques for Heart Disease Prediction from Heart Sound Data

Limited quantity of heart sound data obstacles leveraging the superior performance of deep neural networks. In this work, upon applying effective preprocessing techniques, we first investigate 3 machine learning models to detect 4 multi-label heart diseases and compare their performance with SOTA architecture for audio. Our solution of finetuning audio spectrogram transformer obtained an ICBHI score of 0.6483 and was ranked the runner-up position considering the explored approaches and the intuitions behind those.

Vehicle Object Detection in the Context of Bangladesh Road Traffic

We explore cutting-edge approaches for vehicle object detection under the diverse and challenging road conditions prevalent in Bangladesh. Specifically, we investigate the performance of different YOLO configurations, from YoloV6-L6 to the advanced YoloV8, alongside other novel transformer-based models including MViT, rtDETR and CoDETR.

Supervisor(s):

Tax Wizard: A Personal Income Tax Assistant

This software project implements a tax calculator and payment tracking assistant web application. Building upon basic user interface and backend logic, we develop a CI/CD pipeline to itertively build, test and deploy the solution. Next, we gradually setup load balancing, scaling, monitoring and security scanning features. Finally, we add a visualization and forecasting dashboard to it. Technology stack includes Frontend: NextJS (Javascript), Backend: FastAPI (Python), Database: PostgreSQL, SQLAlchemy and Other Tools: GitHub Actions, Docker, AWS Cloud (ECS, RDS, S3, ELB, CloudWatch), SonarQube, Terraform.

Automating Institutional Course Management System Development

This software project implements a simple web application handling course registration, assignment and result management for students. Building upon basic user interface and backend features, we develop a CI/CD pipeline to itertively build, test and deploy the solution. Next, we gradually setup load balancing, scaling, logging and monitoring features. Technology stack includes Frontend: React (Javascript), Backend: Express (Javascript), Database: PostgreSQL, Prisma and Other Tools: GitHub Actions, Docker, AWS Cloud (Beanstalk, RDS, S3, ELB, CloudWatch).

Machine Learning Algorithms and Neural Network from Scratch

Here we implement matrice transformation, eigen-value decomposition, image reconstruction using singular value decomposition, adaboost algorithm with logistic regression in classfication tasks, exploratory data analysis and preprocessing techniques, training and evaluating a feed forward neural network, principal component analysis and clustering with expectation-maximization algorithm on gaussian mixture models from scratch using only Python3 and Numpy.

Beyond The Seas: A Unified Higher Study Information Portal

This software project addresses the need for a unified platform in pursuit of higher education abroad. It is designed to simplify the application journey for candidates by organizing funding details, offering personalized professor recommendations, facilitating searches based on academic and non-academic preference factors, and employing data analytics and visualization all in one comprehensive platform. Technology stack includes Frontend: ReactJS, NextJS (Javascript), Backend: Flask (Python), Database: PostgreSQL, SQLAlchemy and Other Tools: Git, Docker, AWS Cloud.

Supervisor(s):

Cryptography Algorithms, Security Vulnerabilities, Attacks and Protection

Here we implement public key cryptography system, AES, with Diffie Hellman and RSA key exchange algorithms interconnected via client-server socket. We also explore pedagogical malware (virus and worm) functionalities within docker containers and demonstrate buffer overflow attacks. Finally, we experiment with firewall configurations and demonstrate the features of BeEF, an open-source browser exploitation framwork.

Raster Based Graphics Pipeline and Ray Tracing with OpenGL

We first implement 2D and 3D geometric transformation tasks using OpenGL C++ library. We then implement different steps (modeling, view, projection transformation, clipping, scan conversion) of a raster based graphics rendering pipeline. Finally we implement a ray casting and recursive ray tracing application with OpenGL using phong lighting model that includes a 3D world with objects of various geometric shapes.

Intelligent Search Algorithms and Heuristics To Solve Computationally Complex Puzzles

We leverage prominent heuristics to develop intelligent solvers for different classical problems including n-puzzle solving via A* search, latin square solving by formulating as a constraint satisfaction problem, exam slot scheduling using local search techniques and finally implementing minesweeper using knowledge engineering and first-order inference logic.

Implementing System Calls, Scheduling, Memory Management with xv6

We first explore bash scripting and synchronization using pthreads. Then we implement custom system calls to incorporate command line functionalities and lottery scheduling algorithm into existing xv6 pedagogical operating system. Finally, we implement copy-on-write paging mechanism for xv6 memory management.

VLAN Configuration and Wireless Network Simulation Experiments

We setup Virtual Local Area Network (VLAN) using Cisco Packet Tracer and configure network components (Switch, Router, Trunk, NAT, PAT, ACL). We then use Network Simulator 2 (NS2) to simulate wired and wireless network topologies with varying protocols across different network layers. We also implement a simple multi-threaded file web server using Java socket programming. Finally, we experiment with a modification in Round Trip Time (RTT) and Retransmission Time Out (RTO) estimation algorithms and simulate a black hole security attack on AODV routing protocol.

FireCop: A Miniature Fire Fighting Robot

Here we develop a micro-controller-based low-cost movable robot, facilitated with a minimalistic water supply system to extinguish fire in a small area. Key components include an Arduino Uno as the microcontroller unit and an IR flame detector as the sensory module. We use motor drivers to connect the DC water pump that ejects water via a plastic pipe and the DC gear motors to run the wheels. A servo motor controls the angular rotation of the sensor body and a buzzer module alerts the surroundings upon detecting fire. Finally, rechargeable Li-ion batteries have been used to power up the whole system.

BookKeep: Online Book Store Platform

BookKeep is an online book store platform, primarily focusing on buying and borrowing books online. We implement the important features of a Relational Database Management System (RDBMS) from scratch in this project. Technology stack includes Frontend: HTML5, CSS3, Javascript, JQuery, Bootstrap, Backend: Django (Python), Database: Oracle, SQL, PL/SQL and Other Tools: Git, Docker.

Education

Find below a brief overview of my academic journey (so far).

Work Experience

My goal is to pursue a research-oriented career, be it in academia or industry. Here are some of my professional experiences.

Honors and Awards

I have been blessed with some accolades time to time, thanks to the amazing mentors and peers from whom I learnt a lot.

Research and Travel Grants

Academic Scholarships

• Dean's List Award and University Merit Scholarship:

  Scholarship awarded by BUET for outstanding academic performance in two out of four academic levels of undergraduate studies [2023, 2021]

• NDC Honorable Mention:

  Award of merit (given to top 0.5%) by Notre Dame College, Dhaka for distinction in both academic and extra-curricular activities [2018]

• HSC and SSC Talentpool Scholarship:

  Government scholarship awarded by Dhaka Education Board for brilliant academic performance in Higher Secondary and Secondary School Certificate public exams [2018, 2016]

Hackathon Prizes

• BUET BME Fest BioMed Datathon:

  Runner up in the multi-label heart disease classification competition from heart sound data [2024]

• DU IIT Fest HackTheVerse:

  Champion in the DevOps category of software development competition [2023]

• BUET CSE Fest Hackathon:

  Champion in the DevOps category of software development competition [2022]

Other Recognitions

• অন্তঃবুয়েট বিজয় দিবস আবৃত্তি প্রতিযোগিতা:

  Runner-up medal for oral recitation performance of poetry [2019]

• জাতীয় ভাষা প্রতিযোগ:

  Champion of the largest national contest on Bengali literature and linguistics [2018]

• Josephite Film and Drama Group National Fest:

  Champion of the group drama performance segment [2017]

Leadership

I dedicate a portion of my time to engage in technical and cultural community services. Take a look at selected contributions.

Last updated: November 20, 2024
I found this cool template builds the webpages of some even cooler scientists I know. Thanks to Leonid for implementing Jon's design in Jekyll.