Ninad Nobo

Available for AI/ML Research & Engineering Opportunities

Mohammad Ninad Mahmud Nobo

AI Systems Researcher & Full-Stack Engineer

Focused on LLM Applications, Medical AI, and Software Automation

Dhaka, Bangladesh

BUET CSE

Building reliable AI-driven systems combining LLMs, backend engineering, and applied machine learning for healthcare, automation, and real-world deployment.

AI Systems EngineeringFull-Stack & BackendLLMs & Medical AI

GitHub LinkedIn Google Scholar View CV View Biodata

About

Assalamu Alaikum! I am Mohammad Ninad Mahmud Nobo.

I have recently graduated from Bangladesh University of Engineering and Technology(BUET) with a degree in Computer Science and Engineering. My work focuses on building AI systems that integrate large language models into real-world applications. My work focuses on LLM-based software testing and medical AI with an emphasis on reliability and deployment.

What I bring

I connect AI models to production systems.
I design for reliability in real-world environments.
I move from research ideas to deployable systems.

AI Systems Engineering

Designing end-to-end AI systems integrating LLMs with backend services and real-world deployment pipelines.

Large Language Models

Applying LLMs for reasoning, automation, and system-level applications.

LLM-based Testing

Building automated testing pipelines using LLMs for improved coverage and reliability.

Medical & Trustworthy AI

Developing reliable AI systems for healthcare with conflict-aware reasoning.

Education

School and College

Uttara High School and College

2008 - 2012

PSC (2012): GPA 5

Rajuk Uttara Model College

2013 - 2021

JSC (2015): GPA 5, Talentpool Scholarship

SSC (2018): GPA 5

HSC (2020): GPA 5, General Grade Scholarship

Undergraduate

Bangladesh University of Engineering and Technology (BUET)

BSc in Computer Science and Engineering

2022 - 2026CGPA 3.60 / 4.00

Thesis: AutoTestGenX: Multi-Agent LLM Framework for End-to-End Web Testing

Relevant Coursework

Data Structures & AlgorithmsDigital Logic DesignDatabase SystemsOperating SystemsComputer ArchitectureComputer NetworksComputer GraphicsAlgorithm EngineeringArtificial IntelligenceMachine Learning

Research Interest

Large Language Model SystemsSoftware Testing AutomationApplied Machine LearningMedical AI

Projects

Selected Engineering Projects

View all on GitHub

Featured Project2025AI/MLFull-Stack

MindTrace

View Code (GitHub)

AI-powered dementia care platform enabling real-time caregiver support through an accessibility-first mobile and backend system.

Problem

Caregiver support is fragmented and delayed due to lack of real-time, coordinated tools.

Solution

Built an end-to-end Android + Spring system with AI-assisted workflows and scalable backend services.

Impact

Delivered a real-time healthcare support system with working product and infrastructure demos.

- Developed full-stack Android + backend architecture for real-time support.
- Designed accessibility-first UX tailored for elderly users.
- Integrated AI-assisted workflows with scalable backend services.

KotlinSpring BootSpring AIPostgreSQLRedisFirebaseDockerAzure

Architecture

Android client for caregiver-patient interaction

Spring Boot API orchestrating AI workflows

PostgreSQL + Redis + Firebase for storage and real-time updates

Feature Demo

Infrastructure Demo

Featured Project2025AI/ML

Gemma VetCare

View Code (GitHub)

AI-assisted veterinary decision-support system designed for reliable use in low-connectivity environments.

Problem

Rural livestock care suffers from limited veterinary access and unstable connectivity.

Solution

Built Android + backend decision-support system with resilient APIs and offline-first workflows.

Impact

Enabled reliable AI-assisted guidance in low-connectivity field conditions.

- Developed AI-assisted mobile workflows for livestock decision support.
- Optimized backend APIs for intermittent connectivity.
- Designed robust data flow for real-world field usage.

KotlinSpring BootSpring AIRoomSQLite

Architecture

Offline-first Android workflow

Spring AI inference endpoints

SQLite-based local storage with sync handling

Feature Demo

Featured Project2024Systems

Compiler Construction

View Code (GitHub)

Full compiler for a C-like language with lexical analysis, parsing, semantic checks, and optimized 8086 code generation.

Problem

Compiler stages require precise coordination across parsing, semantics, and code generation.

Solution

Implemented full compiler pipeline using Flex/Bison with optimized 8086 output.

Impact

Demonstrated complete systems-level understanding through a working language toolchain.

- Built full compiler pipeline from tokenization to code generation.
- Implemented semantic analysis and intermediate code generation.

C++17FlexBison8086 CodegenLinux

Architecture

Lexer and parser using Flex/Bison

Semantic analysis and IR generation

Optimized 8086 code emission

Featured Project2025Systems

Computer Graphics Pipeline

View Code (GitHub)

End-to-end graphics pipeline with transformations, rasterization, Z-buffering, and ray tracing using OpenGL.

Problem

Rendering realistic 3D scenes requires coordinated math-intensive pipeline stages.

Solution

Implemented full graphics pipeline with OpenGL visualization and ray tracing support.

Impact

Validated both theoretical and practical understanding of graphics systems.

- Implemented transformations, clipping, and rasterization stages.
- Built Z-buffer and ray tracing for 3D rendering.
- Developed OpenGL demos with interactive camera control.

C++17OpenGLRasterizationRay Tracing

Architecture

Transformation and camera pipeline

Rasterization with Z-buffering

Ray tracing for lighting and reflections

Research

Research and Thesis Work

I design and evaluate AI systems for software testing and medical decision support, with an emphasis on reliability, robustness, and practical impact.

AutoTestGenX: Multi-Agent LLM Framework for End-to-End Web Testing

View Code (GitHub)

Undergraduate Thesis2025 - 2026Ongoing

Multi-agent LLM framework for automated web test generation, verification, and execution from natural-language requirements.

Problem

Manual web test creation is time-consuming and difficult to scale from evolving natural-language requirements.

Solution

Developed a multi-agent LLM framework that automatically generates, verifies, and executes web test suites from functional descriptions.

Impact

Achieved 84.0% test scenario coverage and 90% error detection, outperforming zero-shot and few-shot prompting approaches.

- Developed a multi-agent LLM framework for automated web test generation and execution from natural-language requirements.
- Built benchmark datasets and evaluated generated test suites against engineer-validated ground-truth test suites.
- Achieved 84.0% test scenario coverage and 90% error detection, outperforming zero-shot and few-shot prompting approaches.

Tech Stack

PythonLLM EngineeringMulti-Agent SystemsPlaywrightBrowser Automation

MedCAR: Conflict-Aware Medical Reasoning

View Code (GitHub)

Research Project2026Under Review

Conflict-aware multi-model reasoning system for chest X-ray analysis with confidence-calibrated clinical support.

Problem

Conflicting predictions across medical AI models create uncertainty and reduce reliability in clinical decision-making.

Solution

Designed a reasoning layer with semantic reconciliation and confidence-calibrated abstention.

Impact

Improves trust and safety by converting conflicting outputs into auditable, confidence-aware recommendations.

- Integrated multiple AI models into a unified diagnostic pipeline.
- Developed semantic conflict-resolution with confidence calibration.
- Designed abstention mechanisms for safer decision support.

Tech Stack

PythonPyTorchNumPy

Bengali-Loop: Community Benchmarks for Long-Form Bangla ASR and Speaker Diarization

View Paper (arXiv: 2602.14291)

Preprint2026Preprint (arXiv: 2602.14291)

Large-scale benchmark dataset and evaluation framework for long-form Bangla ASR and speaker diarization.

Problem

Bangla lacks standardized large-scale datasets and evaluation benchmarks for long-form speech recognition and multi-speaker diarization.

Solution

Contributed to building a reproducible benchmark with curated datasets, annotation pipelines, and standardized evaluation protocols.

Impact

Enables consistent benchmarking for Bangla ASR and diarization using Word Error Rate (WER) and Diarization Error Rate (DER).

- Contributed to the development of a benchmark dataset for long-form Bangla ASR and speaker diarization.
- Built data collection and preprocessing pipelines, including subtitle extraction and annotation workflows.
- Supported evaluation using Word Error Rate (WER) and Diarization Error Rate (DER).

Tech Stack

PythonSpeech ProcessingMachine Learning