About me

Assalamu Alaikum,

Hi, I’m MD Sadik Hossain Shanto, a Computer Science and Engineering researcher and AI engineer. I earned my B.Sc. in CSE from Bangladesh University of Engineering and Technology (BUET) and am now pursuing my M.Sc. in CSE there.

My research sits at the intersection of trustworthy and applied AI. I work on the privacy and security of large language and vision-language models, uncertainty in VLMs, and clinical-grade healthcare AI, with seven papers across venues such as ACL, EACL, and Interspeech, alongside ongoing work under review at ACM CCS and Expert Systems with Applications.

I am equally an engineer who ships. As a Remote AI Research Engineer for the Ministry of Defense Healthcare Services in Saudi Arabia, I research, build, and deploy real systems. I architected Digital Harf, a multimodal AI speech and language therapy platform, and Harf-Speech for Arabic phoneme-level assessment, now integrated into Sehhatek, the national healthcare app serving millions of users.

Teaching and mentorship matter deeply to me. Over three years I taught programming and core CS to more than 7,000 students at ROOTs Edu, and I love turning hard ideas into something a beginner can hold onto. I am a strong believer in teamwork and leadership, and I have led teams to national and international recognition in research and cybersecurity.

I believe in pushing limits and embracing every challenge as an opportunity to grow. My life motto is simple: 'Be better than yesterday.' Every day, I aim to learn something new, whether it's a technical skill or a life lesson. I’m driven by the belief that consistent effort, punctuality, and a structured routine are essential to success.

Skills

Languages

C
C++
Java
Python
Javascript
Assembly
Bash
Arduino



Tools

Git
Github
Pytorch
Postman
Swagger
Linux
Kaggle
Huggingface
Xv6
Logisim
AtMega32
NS3
Autopsy
Wireshark
OpenGL




Backend

NodeJS
ExpressJS



Frontend

ReactJs
EJS
Bootstrap
Antdesign
JavaFX
HTML



Data Systems

PostgreSQL
OracleDB
Prisma
MySQL



Clouds

Digital



Education

  1. Bangladesh University of Engineering and Technology (BUET)

    June 2025 - Present
    M.Sc. in Computer Science and Engineering
    CGPA: 4.00/4.00 (9 credits completed)

    Pursuing graduate study and research in Computer Science and Engineering.

  2. Bangladesh University of Engineering and Technology (BUET)

    February 2020 - March 2025
    B.Sc. in Computer Science and Engineering
    CGPA: 3.90/4.00

    Three-time Dean's List Award & Two-time University Merit recipient.

    Notable Courses:

    • CSE 107 - Object Oriented Programming

    • CSE 215 - Database

    • CSE 203 & 207 - Data Structures and Algorithms

    • CSE 205 - Digital Logic Design

    • CSE 305 - Computer Architecture

    • CSE 307 - Software Engineering

    • CSE 309 - Compiler Design

    • CSE 313 - Operating Systems

    • CSE 317 - Artificial Intelligence

    • CSE 321 - Computer Networks

    • CSE 325 - Information System Design

    • CSE 405 - Computer Security

    • CSE 409 - Computer Graphics

    • CSE 463 - Introduction to Bioinformatics

    • CSE 471 - Machine Learning

    • CSE 411 - Simulation and Modeling

  3. Rajshahi College, Rajshahi, Bangladesh

    2019
    Higher Secondary Certificate (HSC)
    GPA: 5.00/5.00

    Received board scholarship (Talent Pool).

  4. Govt. Laboratory High School, Rajshahi, Bangladesh

    2017
    Secondary School Certificate (SSC)
    GPA: 5.00/5.00

    Received board scholarship (Talent Pool).

Experience

Ministry of Defense, Healthcare Services

Lead: Dr. Ehsan Hoque (Professor, University of Rochester; MIT PhD), with Sabri Boughorbel (QCRI) and Abdulrhman Aljouie (KSAU-HS).

  1. Remote AI Research Engineer

    April 2025 - May 2026
    Remote · Kingdom of Saudi Arabia

    Research, build, and deploy clinical-grade AI for healthcare and autism (ASD) speech & language therapy. Two papers (Interspeech 2026 + preprint) emerged from this work.


    • Architected Digital Harf, a multimodal AI therapy platform (3 therapy modules + an AI Caregiver Assistant) powered by an Agentic Synthetic Data Engine that auto-generates culturally grounded content.

    • Built Harf-Speech, fine-tuning Whisper, Omni-ASR, and wav2vec2 for Arabic phoneme-level assessment with advantages over proprietary Azure Cognitive Services.

    • Integrated into Sehhatek, Saudi Arabia's national healthcare app serving millions of users (in staging, clinical trial beginning), with partners King Fahad Armed Forces Hospital, Mohammed Bin Salman Autism Program & Madinah Autism Society.

    • Contributed to cross-team efforts: Waee (autism screening, University of Cambridge collaboration) and AIDA (autism screening, MIT Media Lab collaboration).


ROOTs Edu

  1. Online ICT & Programming Instructor

    June 2022 - January 2025
    Remote

    • Taught ICT, introductory programming and CS basics to 7,000+ students over three years at ROOTs Edu, one of Bangladesh's leading online EdTech platforms, covering languages, data structures, algorithms, and OOP fundamentals.

    • Developed structured lesson plans and assessments from scratch, translating abstract CS concepts into accessible, beginner-friendly modules.


Others

  1. Coding Expert

    June 2023 - December 2023
    Remotasks.com · Scale AI · Remote

    • Trained and evaluated the responses, code generation, completion and debugging capabilities of popular LLMs like GPT-4 and Gemini to improve their coding capabilities.

    • Joined as a coding expert on the 'Dolphin Coding' project, solving programming problems and providing solutions with proper documentation and explanations that were later fed into the LLMs.


Research & Publications

  1. 1
    Published ACL 2025 (Student Research Workshop) 22 Citations

    SequentialBreak: Large Language Models Can be Fooled by Embedding Jailbreak Prompts into Sequential Prompt Chains

    with Md. Rafi Ur Rashid, PhD Student, Pennsylvania State University, USA

    As large language models move into real products, their safety guardrails stay surprisingly brittle. SequentialBreak exposes a striking blind spot: burying a harmful request inside a chain of ordinary ones slips it past these defenses in a single query, revealing how fragile the alignment really is.

    AI Safety & SecurityLarge Language Models
    Paper
  2. 2
    Published EACL 2026 (Findings)

    The Art of Saying “Maybe”: A Conformal Lens for Uncertainty Benchmarking in VLMs

    with Dr. Mohammad Saifur Rahman, Professor, BUET, Md. Rizwan Parvez, Research Scientist, QCRI

    Today's vision-language models stay confident even when they are wrong. This work tackles a neglected question, do they know what they don't know, delivering the first broad benchmark of their uncertainty and a way to measure it even for closed, commercial models.

    VLM UncertaintyVision-Language Models
    Paper
  3. 3
    Published Interspeech 2026

    Harf-Speech: A Clinically Aligned Framework for Arabic Phoneme-Level Speech Assessment

    with Dr. Ehsan Hoque, Professor, University of Rochester (MIT PhD), Sabri Boughorbel, QCRI, Abdulrhman Aljouie, KSAU-HS

    Remote AI Research Engineer, Ministry of Defense Healthcare Services, Saudi Arabia

    What if speech assessment were available anytime, anywhere, without losing clinical quality? With Arabic-trained speech pathologists scarce, Harf-Speech automatically evaluates Arabic (MSA) at the same fine-grained level clinicians use, with results validated against certified specialists.

    Healthcare AISpeechHuman-AI Interaction
    Paper
  4. 4
    Under Review

    Digital Harf: A Clinically Integrated Multimodal AI System for Pervasive Arabic Speech and Language Therapy

    with Dr. Ehsan Hoque, Professor, University of Rochester (MIT PhD), Sabri Boughorbel, QCRI, Abdulrhman Aljouie, KSAU-HS

    Remote AI Research Engineer, Ministry of Defense Healthcare Services, Saudi Arabia

    Children with autism in Arabic-speaking regions face a shortage of specialists and almost no culturally relevant therapy material. Digital Harf carries clinician-led speech therapy into the home through an AI engine that generates culturally grounded content, treating cultural fit as core infrastructure rather than an afterthought.

    Healthcare AIAgentic SystemsMultimodalityHuman-AI Interaction
    Preprint
  5. 5
    Under Review ACM CCS 2026

    Robustness of Vision Language Models Against Split-Image Harmful Input Attacks

    with Dr. Shagufta Mehnaz, Assistant Professor, Pennsylvania State University, USA, Md. Rafi Ur Rashid, PhD Student, Pennsylvania State University, USA, Vishnu Dasu, PhD Student, Pennsylvania State University, USA

    Vision-language models have learned to refuse harmful images, but only when the harm arrives in one piece. This work reveals that splitting a harmful image across fragments quietly defeats their safety training, introduces attacks that expose this overlooked gap, and proposes a defense to close it.

    AI Safety & SecurityVision-Language Models
    Preprint
  6. 6
    Published IEEE Signal Processing Cup at ICASSP 2025 1st Runner-Up Team Lead 2 Citations

    DFCon: Attention-Driven Supervised Contrastive Learning for Robust Deepfake Detection

    with Dr. Mohammad Saifur Rahman, Professor, BUET

    As AI-generated faces grow indistinguishable from real ones, catching them in the wild remains an open problem. Leading this BUET team at IEEE ICASSP 2025, we built a deepfake detector that blends complementary vision models to generalize across unseen generation methods and source datasets.

    Computer VisionDeepfake Detection
    CompetitionReport
  7. 7
    Under Review Expert Systems with Applications (Journal) Highly Commended, Global Undergraduate Awards

    AttMetNet: Attention-Enhanced Deep Neural Network for Methane Plume Detection in Sentinel-2 Satellite Imagery

    with Dr. Tanzima Hashem, Professor, BUET, Sultanul Arifin, Lecturer, BRAC University, Dhaka

    Methane is a potent driver of climate change, yet spotting its plumes from satellites is buried under false alarms that demand expert review. AttMetNet is a lightweight, attention-driven model that hones in on a methane-sensitive spectral signature to flag true plumes, making reliable global-scale monitoring practical.

    Remote SensingComputer Vision
    AwardPreprint

Development & Other Projects

  • ReMan - Retailers and Manufacturers

    ReMan - Retailers and Manufacturers

    🏷️ Nodejs React PostgreSQL Prisma Antd Swagger

    Here we tried to connect The Retailers and Manufacturers through a completely digital platform. The platform provides features such as inventory and bulk order management, production management, data analytics and market analysis to help retailers and manufacturers streamline their operations and maintain a efficinet supply chain.

  • Health++

    Health++

    🏷️ Nodejs Ejs Bootstrap PostgreSQL

    A digital platform for the health sector. This project aims to streamline healthcare services by providing an integrated system for patient management, appointment scheduling, centralized medical records management, digital prescription with medicine suggestions, centralized test result upload and validation system etc. It has seperate modules for patients, doctors, lab assistant and admins.

  • Pacman

    Pacman

    🏷️ C iGraphics

    Built a pc version of the famous game Pacman with c and iGraphics. Enjoy the game with inky, blinky, pinky and clyde.

  • BeEF - Browser Exploitation Framework Tool

    BeEF - Browser Exploitation Framework Tool

    🏷️ BeEF Security and Pentesting

    Here we investigated BeEF: The Browser Exploitation Framework, an influential and open source security tool designed for browser penetration testing. We successfully extended some of the modules and demonstrated the security vulnerabilities with some live attacks.

  • Intra BUET CTFd

    Intra BUET CTFd

    🏷️ DigitalOcean Droplets Load Balancer Redis Cache

    Helped setting up the infrastructure and hosting services for Intra BUET CTF - 2024

  • 6 DOF Robotic Arm Mounted on a bot

    6 DOF Robotic Arm Mounted on a bot

    🏷️ Microcontrollers Arduino Mega Servo and DC motors Bluetooth Module

    Built a 6 degree of freedom robotic arm with a mobile bot. Used components: Arduino Mega, 6 servo motors, 4 DC motors, 1 bluetooth module, LiPo battery etc.

  • Yet Another C Compiler

    Yet Another C Compiler

    🏷️ Bison Yacc C

    Built a compiler from scratch with modern optimizations for a subset of C language.

  • Computer Graphics - Ray Tracing

    Computer Graphics - Ray Tracing

    🏷️ C++ OpenGL

    Implemented ray tracing from scratch with OpenGL to render 3D scenes with realistic lighting, shadows, and reflections. Implemented various features such as recursive reflection, and refraction to enhance the visual quality of the rendered images.

  • Xv6 - Operating System - System Call, Scheduler, IPC, Lock , Semaphore and Threading Implementation

    Xv6 - Operating System - System Call, Scheduler, IPC, Lock , Semaphore and Threading Implementation

    🏷️ C XV6 Operating System

    Extended the xv6 operating system by implementing various system calls, process scheduler, inter-process communication mechanisms, locks, semaphores, and threading support.

  • 4 bit MIPS PC

    4 bit MIPS PC

    🏷️ AtMega32 Logisim Computer Architechture

    Built a modified and reduced version of the MIPS instruction set with ATMega32 and other hardware components

Achievements

Research & Competition

  1. IEEE Signal Processing Cup at ICASSP 2025

    1st Runner-Up · Team Lead

    Secured 1st Runner-Up in the 'Deepfake Face Detection in the Wild' challenge at IEEE ICASSP 2025, leading the BUET team.

  2. The Global Undergraduate Awards

    Highly Commended (Computer Science)

    Undergraduate thesis recognized among the top undergraduate research worldwide in the Computer Science category.

Academic

  1. Dean's List Award & University Merit, BUET

    Dean's List (3×) · University Merit (2×)

    Awarded the Dean's List for outstanding academic performance in three academic years, and University Merit for two semesters at BUET.

Capture The Flag (CTF)

  1. Intra BUET CTF 2023

    1st Runner-Up

    Achieved 2nd place among 35 teams in the cybersecurity competition organized by the BUET Cyber Security Club.

  2. BD-Flag Hunt CTF 2022

    9th Position

    Ranked 9th among 100+ teams in this national cybersecurity challenge.

Others

  1. Microcontroller, Microprocessor and Embedded System Sessional Project

    Honorable Mention

    A 6-DOF robotic arm mounted on a bot, awarded Honorable Mention in the CSE Department, BUET.

Contributions

Leadership & Contributions

  1. Deep Learning Sprint, BUET CSE FEST 2025

    2025
    Organizer

    Problem formulation, dataset curation & management for a 120+ team open deep learning contest at BUET's flagship technical festival.

    Part 1Part 2

  2. Capture The Flag (CTF), BUET CSE FEST 2024

    2024
    Organizer

    Organized and maintained the full competition infrastructure for a 60+ team inter-university cybersecurity challenge.

    Event

  3. BUET Cyber Security Club

    2024 - 2025
    Director of Technology

    Led technical operations and organized inter-university cybersecurity competitions and workshops, including hosting and maintaining CTF infrastructure on DigitalOcean.