I am a final-year Computer Science & Engineering student at United International University (UIU), graduating in October 2026. My core passions lie at the intersection of Artificial Intelligence (NLP, Computer Vision & Edge AI), DevSecOps & Cloud Security, and Embedded IoT Systems.

Throughout my academic journey and professional engagements, I have focused on solving high-impact engineering and research challenges — from pioneering hallucination detection in Bengali Large Language Models and deploying TinyML Quantization-Aware Federated Learning (QAF-Net) for clinical neuroimaging, to architecting heterogeneous GNN-powered cloud misconfiguration detection engines and low-cost airborne CanSat telemetry frameworks.

• 🔬 Bengali NLP & LLM Robustness: Researched and benchmarked Factual Hallucination Detection in Generative AI for Bengali Text using fine-tuned transformer models (BanglaBERT and mBERT), curating a 7,043-sample multi-LLM benchmark across GPT, Gemini, and Claude.
• 🧠 Medical Edge AI & Federated Learning: Co-engineered QAF-Net, an ultra-compact (606 KB INT8) TinyML cross-dataset brain tumor MRI classification pipeline utilizing site-local normalization and exposing 55.2% near-duplicate leakage in global Kaggle benchmarks.
• ☁️ Cloud Security & DevSecOps: Designed the Intelligent Cloud Misconfiguration Detection & Automated Remediation framework uniting 64-D AST projections, Heterogeneous Threat Graphs (R-GCN), and calibrated ML-LLM triage for Kubernetes and AWS workloads.
• 🛰️ Embedded & Aerospace IoT: Co-engineered an AI-driven CanSat Microclimate Monitoring Framework integrating multi-sensor arrays, XBee wireless telemetry, and satellite earth observation data fusion.
• 💼 Professional Experience: Hands-on experience as a Data Labelling Associate and Image Annotator at ARC Annotation, specializing in polygon, bounding-box, and semantic segmentation quality assurance.
• 🏛️ Leadership & Community: Former Executive at UIU App Forum (2022–2023), coordinating campus technical seminars, collaborative workshops, and student development initiatives.
• 🎯 Career Interests: Software Engineering, Machine Learning Research, DevSecOps / Cloud Security, and Software Quality Assurance (SQA).

Programming Languages
AI, ML & NLP
Cloud & DevSecOps
Web & Backend
Databases
IoT & Embedded
Testing & SQA
Tools & Platforms

Undergraduate Thesis Project | Deep Learning & Low-Resource NLP
Supervised by Dr. Ohidujjaman, Associate Professor, United International University

• Built the first benchmark dataset of 7,043 manually verified Bengali Q&A pairs generated across OpenAI GPT, Google Gemini, and Anthropic Claude.
• Formulated an empirical audit revealing a 56.5% hallucination occurrence rate in low-resource Bengali LLM generations.
• Fine-tuned domain-specific BanglaBERT (ELECTRA) vs. mBERT, achieving 82.47% accuracy, 82.45% Macro F1, and 2.07x inference throughput (1,358 samples/s).
• Tech Stack: Python, PyTorch, Hugging Face Transformers, BanglaBERT, mBERT, Google Colab (Tesla T4)

Quantization-Aware Federated Learning with Site-Local Normalization for TinyML Clinical Neuroimaging
Developed in the Department of Computer Science & Engineering, United International University

• Formulated QAF-Net, uniting post-training INT8 quantization residual regularization and leave-one-site-out federated aggregation across disparate hospital cohorts (China, Bangladesh).
• Uncovered and decontaminated a critical 55.2% near-duplicate leakage in widely used Kaggle benchmarks, proving true cross-scanner generalization.
• Delivered an ultra-compact 606 KB INT8 quantized model fitting in <1 MB SRAM/Flash with 94.7% accuracy and negligible 0.15-point quantization loss.
• Implemented Grad-CAM interpretability to validate anatomical localization of tumor pathology on edge-rendered predictions.
• Tech Stack: Python, TensorFlow / Keras, TensorFlow Lite (INT8), Federated Learning (FedAvg/FedProx/FedBN), OpenCV, NumPy, LaTeX

Multi-Domain DevSecOps Framework: AST Feature Extraction, Heterogeneous GNNs & Calibrated ML-LLM Triage
Complete IEEE Camera-Ready Publication Assets, Overleaf Package & Defense Viva Guide

• Dense 64-D AST Feature Engine: Projects heterogeneous YAML/JSON manifests into a unified numerical feature space modeling privilege boundaries, RBAC bindings, and IAM trust scopes.
• Relational Threat Graph (R-GCN): Unifies decoupled manifests into a directed threat graph to detect multi-hop lateral privilege escalation chains missed by traditional linters.
• Cost-Optimized ML-LLM Triage: Sigmoid-calibrated Random Forest resolves $>90%$ of manifests locally in $<2\text{ ms}$, escalating only uncertain edge cases to LLM Chain-of-Thought reasoning (slashing API costs by 92%).
• Closed-Loop Sandbox Remediation: Pairs TreeSHAP explainability with compiler verification sandboxing to guarantee 100% syntactically valid YAML automated fixes.
• Tech Stack: Python 3.10+, PyTorch Geometric (R-GCN), Scikit-Learn, TreeSHAP, YAML / AST, Kubernetes, AWS IAM / SAM, LaTeX (IEEEtran)

Aerospace Sensing, Telemetry & Multi-Source Data Fusion Framework

• Designed an airborne CanSat payload equipped with BME680, MQ-135, GPS, and a 6-DOF IMU driven by a Teensy 4.1 / ESP32 controller.
• Engineered point-to-point XBee wireless telemetry with 94%–98% Packet Delivery Ratio (PDR) up to 1 km.
• Implemented a multi-source feature fusion pipeline merging airborne telemetry with macroscopic satellite earth-observation data (LST, AOD, NDVI).
• Integrated an edge Isolation Forest model detecting environmental anomalies with 84.7% accuracy at ultra-low inference latency (18–25 ms).
• Tech Stack: Embedded C/C++, Teensy 4.1, ESP32, XBee RF, Python, Scikit-Learn, LaTeX (IEEE Format)

Enterprise-Grade Full-Stack Management & Organizational Coordination Suite

• AI Task Decomposition: Integrates Groq Cloud LLaMA-3.1-8B to break down massive deliverables into actionable numbered subtasks in under 120 words.
• Skill-Based Allocation Engine: Algorithmic task recommendation matching employee technical proficiencies (beginner to expert) and active workload queues.
• Full Agile Observability: Responsive 5-column drag-and-drop Kanban board, interactive Gantt charts, dynamic multi-tier organization tree visualizer, and an SOS developer distress escalation pipeline.
• Tech Stack: PHP 8.2 (OOP & PDO), MySQL, Groq Cloud API (LLaMA-3.1), JavaScript (ES6+), PHPMailer (SMTP)

Integrated Multi-Safety Embedded IoT Prototype for Senior Care
Developed by Team Micro Minds | Microprocessor & Microcontroller Lab

• Automated Medicine Cabinet: DS3231 RTC synchronized SG90 servo latch with 5–10 min dosage compliance windows and missed-dose alerts.
• Kitchen Hazard Mitigation: MQ-2 gas/smoke sensor and flame detector triggering automatic gas shut-off valve closure and DC exhaust ventilation via relay.
• Smart Security: RC522 RFID door access with ESP32-CAM (OV2640) intruder snapshot capture on repeated unauthorized attempts.
• Fall Detection & SOS: MPU-6050 kinematic fall algorithm (free-fall weightlessness + impact surge + inactivity window) with tactile manual SOS button.
• 📺 Watch Hardware Prototype Video Demonstrations on Google Drive
• Tech Stack: Arduino UNO R3, ESP32 DevKit, ESP32-CAM, MPU-6050, DS3231, MQ-2, Embedded C++

Advanced Object-Oriented Programming (AOOP) Capstone Web Platform

• Designed for academic researchers, supervisors, and students to coordinate proposals, project milestones, and paper submissions.
• Built with Spring Boot, Spring Data JPA, and Spring Security with role-based access control and persistent relational storage.
• Tech Stack: Java, Spring Boot, Spring Security, PostgreSQL / MySQL, Thymeleaf, JavaScript

Urban Mobility & Shared Economy Parking Platform

• Connects vehicle drivers seeking parking spots with local property owners having vacant spaces.
• Designed high-fidelity prototypes in Figma and implemented interactive web interfaces with role-based navigation and booking workflows.
• Tech Stack: HTML5, CSS3, JavaScript, Figma (UI/UX Prototyping)

Campus Accommodation Discovery & Peer Roommate Matching Web Platform

• Tailored for university students relocating to Madani Avenue near UIU campus.
• Features dual-role onboarding (Students vs. Landlords), detailed apartment amenity audits (CCTV, lift, tiled washrooms, service charges in BDT), and dark/light mode toggles.
• Implements an academic peer-matching board filtering by department, completed trimesters, and living habits.
• Tech Stack: HTML5, CSS3 (Flexbox/Grid), JavaScript (ES6+), Bootstrap 4, SVG Animations

• UX Research Strategy with AI — Grameenphone Academy
• Smart CV — Grameenphone Academy (2026)
• Art of Communication — Grameenphone Academy (2026)
• Corporate Presentation Skills — Grameenphone Academy (2026)
• CCNA: Introduction to Networks — Cisco Networking Academy (2023)

I am always eager to collaborate on innovative open-source projects, academic research, and engineering opportunities. Feel free to connect!

• 📍 Location: Dhaka, Bangladesh
• 📧 Email: robiulislamtamim12345@gmail.com
• 📞 Phone / WhatsApp: +880 1970656185
• 🌐 GitHub: github.com/robiultamim
• 🗣️ Languages: Bangla (Native), English (Professional Working Proficiency)

_{Designed & Developed with care by Robiul Islam Tamim • © 2026 All Rights Reserved}