CV
Education
- Bachelor of Science in Computer Science and Engineering, Bangladesh University of Engineering and Technology (BUET), February 2020 – March 2025
Research Experience
- Prof. Reyhaneh Jabbarvand’s Group, Intelligent CAT Lab, University of Illinois Urbana-Champaign (June 2025 – Present)
- Designed a novel agentic framework for automated generation of behavior-driven unit tests for C programs using AI Agents, ensuring comprehensive coverage of functional scenarios
- Proposed an agent decomposition strategy, designing a dual-phase framework that decouples high-level reasoning about program behavior (System 2) from concrete test code generation (System 1), addressing limitations of monolithic AI Agents
- Implemented a neuro-symbolic algorithm to generate program control flow representations in the reasoning phase, isolating model understanding from surface-level generation errors and achieving 2× improvement in code coverage over standard monolithic agents
- Evaluated advanced testing tools including KLEE for symbolic execution and AFL++ for fuzzing, identifying their limitations in test generation
- Prof. Mahmuda Naznin’s Group, BUET (March 2024 – Present)
- Developed ATLAS, a novel static analysis tool addressing limitations of existing single-function program analysis tools by supporting whole-project scope for C and C++ programs
- Designed algorithms for constructing statement-level Control Flow Graphs (CFGs) and Data Flow Graphs (DFGs) with support for inter-procedural analysis across function boundaries
- Extended analysis capabilities from single-function to whole-project scope including multi-file support using Tree-sitter, enabling AST/CFG/DFG generation for large-scale C/C++ codebases
- Developed a Generative Adversarial Network (GAN) architecture to distinguish between legitimate and malicious user profiles injected through push and nuke attack strategies in recommender systems
- Evaluated detection performance on benchmark datasets (MovieLens, Amazon Sales), demonstrating improved precision and recall over traditional unsupervised learning defense mechanisms
- Developed ATLAS, a novel static analysis tool addressing limitations of existing single-function program analysis tools by supporting whole-project scope for C and C++ programs
- Prof. Anindya Iqbal’s Group, BUET - World Bank-Bangladesh Government Collaboration (March 2025 – Present)
- Developed Trace2Script, an AI Agent-powered framework that translates natural language test cases into executable and robust Playwright test scripts for automated Web UI testing
- Proposed and engineered the Web Agent architecture for simulating realistic human actions during autonomous Web UI traversal with comprehensive trace logging of element interactions
- Constructed and validated a comprehensive benchmark dataset of 445 natural language test cases spanning 19 production web applications, ranging from basic UI interactions to complex workflow scenarios
- Demonstrated that software-aware constraints with explicit behavior modeling significantly outperformed unconstrained LLM-based test script generation in evaluation across production applications
- Developed a full-stack inference software for on-premise deployment of fine-tuned LLMs, serving 300 users
- Fine-tuned 7B-parameter models (Llama3, Mistral) using LoRA-based optimization, achieving SOTA model performance for software test case generation from business requirement documents
- Developed Trace2Script, an AI Agent-powered framework that translates natural language test cases into executable and robust Playwright test scripts for automated Web UI testing
Industry Experience
- Software Engineer, Yobo AI, Ontario, Canada (June 2024 – Present)
- Developed a voice-based conversational AI Agent for automating restaurant operations (delivery, pickup, reservation, FAQs), deployed across 27 restaurants
- Engineered solutions for multilingual voice interactions (English, Bengali, Hindi), addressing transcription errors and language pragmatics challenges in production environments
- Led the design of a dynamic context injection pipeline that retrieves and integrates specific knowledge into LLMs, enabling efficient querying of large-scale knowledge bases (restaurant menus, business logic)
- Engineered the end-to-end delivery feature, including address capturing, validation using Google Maps API, and automated delivery acceptance/rejection based on distance calculation and dynamic delivery fee integration
Programming Skills
- Programming Languages: Python, Java, C/C++, TypeScript, JavaScript, SQL, Shell Scripting
- Web Development: React, Flask, FastAPI, Django, Spring Boot, HTML, CSS
- AI/ML & Data: TensorFlow, PyTorch, LangChain, LangGraph, CrewAI, PostgreSQL, SQLite, Supabase
- DevOps & Cloud: Docker, Linux, DigitalOcean
- Tools & Technologies: Git, GitHub, NodeJS, Selenium, AFL++, KLEE, Logstash, Hyperledger Fabric
Manuscript Submitted/Under Preparation
- ATLAS: Automated Tree-based Language Analysis System for C and C++ source programs
- Dual-Phase Agent Decomposition for Automated Unit Test Generation in C Programs
- Under Preparation
- J. M. Chowdhury*, F. Oscar*, & R. Jabbarvand
- Trace2Script: Framework for Web UI Test Script Generation from Natural Language Test Cases
- Under Preparation
- J. M. Chowdhury*, S. M. R. Pial*, J. Y. Khan, M. R. Z. Ratul, & A. Iqbal
* denotes equal contribution
Achievements & Awards
- Bangladesh Blockchain Olympiad 2024 - September 2024
- Honourable Mention in Both AI Category & Blockchain Category
- Led team to Honourable Mention among 250+ teams by developing an AI-powered health ecosystem featuring machine learning medical diagnosis and empathic AI therapist
- Designed MEDIBLOC, a blockchain-based medical prescription management system with Attribute-Based Access Control (ABAC) for secure sharing of prescriptions between hospitals, pharmacies, and patients
- Hello Yobo AI Hackathon 2024 - May 2024
- Second Runners up
- Led team to 2nd runners-up position among 50+ teams by developing Athene, an offline Android LLM chatbot using Mediapipe API with 600 tokens/sec inference speed on mobile CPUs
- Therap Javafest 2023 - August 2023
- Top 10 Finalist
- Led team to Grand Finale (top 10/150+ teams) by developing a fullstack customer management application using Spring Boot and Angular with AI-powered analytics for revenue tracking and forecasting
- SUST Technovent 2023 - January 2023
- First Runners up
- Led team to 2nd runners-up position among 170+ teams by developing a blockchain-based authentication system with RFID technology for vault security. Custom JavaScript blockchain implementation passed 69 test cases
- HackNSU Season 3 - November 2021
- Second Runners up
- Led team to 2nd runners-up position among 30+ teams by developing Remedi, a telemedicine platform with AI-powered symptom diagnosis chatbot using Intelligent Machines API
- Academic Olympiads - 2017-2020
- Physics, Chemistry, Mathematics
- Mathematics: First place winner at Jahangirnagar University Maths Olympiad
- Physics: Translated official Olympiad book (Categories A & B) from Bangla to English; National Round participant (2017-2019)
- Chemistry: Achieved 11th place in National Round competition
