2020
COMPLETED01
College Completed
Academic foundation
Completed higher secondary education and prepared for the next stage in computer science.
01 AI ENGINEER · SOFTWARE BUILDER · RESEARCHER
I build AI
that moves beyond
the notebook.
I’m Md. Minhajul Islam — a CSE graduate and AI Engineer Intern at NEXVIX, turning computer vision, deep learning, and agentic ideas into tested software products.
01 / 05COMPUTER VISIONSEEING COMPLEX SCENES
MOVE TO BEND THE FIELD
SCROLL TO EXPLORE
Machine LearningMODELS THAT LEARN
Deep LearningVISION + GENERATION
Software SystemsBUILT TO OPERATE
ResearchMEASURED, REPRODUCIBLE
02 ENGINEERING TOOLKIT
Code, models,
and systems.
Built through production-minded projects, research experiments, and end-to-end software. Each capability now assembles visibly from 0% when it enters view.
Python
0%
PyTorch
0%
Computer Vision
0%
Deep Learning
0%
C++
0%
JavaScript
0%
SQL & Data
0%
Git / GitHub
0%
Docker & Containers
0%
CI/CD Pipelines
0%
03 SELECTED WORK
From problem
to proof.
Each project explains the real problem, the system I built, and the engineering evidence behind it.
01COMPUTER VISION RESEARCH
PhantomFill
Removes people from crowded photographs and reconstructs the hidden background. I built a multi-stage pipeline that detects each person, refines the mask, performs global inpainting, and locally restores difficult regions.
YOLOv8x-seg → LaMa → Stable Diffusion 2, evaluated for visual quality and residual detections.
PyTorchOpenCVDiffusers
02AI SYSTEMS ENGINEERING
NexDrive AI
Turns a dealership into an AI-assisted operation. Customers can search inventory and speak with a voice agent while staff manage leads, appointments, inventory, and traceable call outcomes.
Production-style FastAPI architecture with secure Supabase data access and deterministic relational test data.
FastAPISupabaseVoice AI
03GENERATIVE AI / THESIS
Lightweight Diffusion
Investigates how to make image inpainting faster without sacrificing generation quality. I designed and compared lighter U-Net configurations against a Stable Diffusion baseline.
Eight-month senior design project with reproducible experiments, deployment code, and quality-efficiency analysis.
U-NetCelebA + COCOGradio
04RESPONSIBLE AI
MindCare
Guides a user through a PHQ-9 mental-health screening conversation, calculates severity, detects crisis language, and produces a downloadable assessment report.
Local LLaMA 3 inference keeps the conversational model on-device; JWT and MongoDB support secure user sessions.
OllamaNode.jsMongoDB
05MEDICAL IMAGING
MRI Tumor Classifier
Classifies brain tumors from MRI scans. The work progressed from an overly simple binary dataset to a more realistic multiclass Figshare dataset and a complete CNN evaluation workflow.
Shows model-development judgment: dataset correction, preprocessing, multiclass training, and evidence-based evaluation.
CNNMRI ImagingJupyter
06SOFTWARE ENGINEERING
Chat-Wave
A browser-based communication product where authenticated users exchange messages instantly through persistent, bidirectional connections.
Built as a software-engineering project with Express APIs, Socket.io events, and a responsive Tailwind/DaisyUI interface.
Express.jsSocket.ioDaisyUI
07AGENTIC PRODUCT ENGINEERING
Northstar Voice Agent
Connects a conversational sales agent to live vehicle inventory and real appointment availability. The agent can answer product questions, explain trade-offs, find matching cars, and schedule a test drive during the call.
Containerized Node service with protected inventory tools, Cal.com booking integration, health checks, and production deployment guidance.
Grok VoiceREST APIsDocker
08FULL-STACK AUTOMATION
Automatic Toll System
Reduces manual toll-booth processing through a web-based workflow for vehicle identification, transaction handling, and centralized record management.
Demonstrates end-to-end product thinking across interface design, application logic, and persistent operational data.
JavaScriptDatabaseAutomation
09E-COMMERCE / AI RETAIL
NexGen Lifestyle
Brings fashion shopping and customer support into one AI-assisted retail experience. A React storefront connects to tools for product search, comparison, stock checks, order tracking, and return eligibility.
FastAPI + Supabase tool gateway with typed requests, policy retrieval, audit logs, and confirmation checks for order changes—keeping database credentials out of the AI layer.
ReactFastAPISupabase
04 SYSTEMS IN MOTION
See what the
projects do.
A rotating product-level view of the experiences and pipelines behind the repositories.
INPUT + MASK
LATENT RECONSTRUCTION
RECONSTRUCTED
MASK FILLEDstructure → texture → refinement
Find a family SUV under $35kQualified lead · family / safety
INTENTTOOL PLANLIVE INVENTORYRANKED MATCHES
012024 Toyota RAV4$32,900 · 94% matchBest overall fit
022023 Honda CR-V$30,500 · 91% matchStrong alternative
CRM lead createdTest-drive slots readyConversation logged
AI
Avery · Voice AgentConnected · 01:24
“I found two SUVs within your budget. Would you like to book a test drive?”
inventory.search ✓calendar.slots ✓
CW
Chat-Wave · Engineering4 members online · live Socket.io room
● LIVEMIMinhajulAKAyeshaRNRayan◉ 00:24
MRI · AXIAL · class 01 / 04
MODEL OUTPUT
Glioma detected
Confidence94.2%
4-class CNN · Glioma / Meningioma / Pituitary / No tumor
AI
MindCare is listeningPrivate local inference · LLaMA 3“Over the last two weeks, how often have you felt little interest or pleasure in doing things?”
PHQ-9 progress7 / 9
Current score8Moderate range · supportive guidance prepared
LIVE EXPERIMENT SYSTEMVISION SYSTEM
01
UnderstandLocate the visual failure
RUNNING02
ResearchCompare models + evidence
QUEUED03
CategorizeMasks, scenes, edge cases
QUEUED04
PrototypeBuild the hybrid pipeline
QUEUED05
ValidateQuality, speed, residuals
QUEUED06
ShipInteractive final product
QUEUEDOUTCOMEDEPLOYED INPAINTING DEMO
05 RESEARCH PRACTICE
Curiosity,
made measurable.
I approach AI as an engineering discipline: understand the failure, research the evidence, build the smallest convincing system, validate it, and turn the result into something people can use.
Current roleAI Engineer InternNEXVIX · Engineering in progress
Research internshipCompletedImage Segmentation & Object Removal
EducationCSE GraduateNorth South University
01
Google Scholar A Machine Learning-Driven Framework for Enhancing Cognitive Function Using tDCS and Brain Gym Interventions
02
Machine Learning-Enhanced Cardiovascular Disease Risk Prediction: A Clinical Intelligence Framework
FORTHCOMING06 CAREER SIGNAL
A trajectory
built to compound.
Education became research. Research became engineering. The next step is building ambitious AI products with a team that values depth, ownership, and measurable impact.
2022—MAY 2026
COMPLETED02
CSE Graduate
North South University
Built foundations in software engineering, machine learning, deep learning, and computer vision.
2025—2026
COMPLETED03
Applied AI Research
Senior Design + Directed Research
Began university research in lightweight diffusion, image inpainting, and object removal.
2026—MAY 2026
COMPLETED04
Research Intern
Image Segmentation & Object Removal
Advanced a multi-stage computer-vision pipeline from failure analysis to reproducible results.
AUG 2026—NOW
CURRENT05
AI Engineer Intern
NEXVIX
Engineering AI capabilities and dependable software workflows for real product use.
NEXT
OPEN06
AI / Software Engineer
High-impact product team
Ready to build intelligent systems that move from research evidence to production value.
DHAKA, BANGLADESH AI ENGINEER · CSE GRADUATE
07 THE PERSON BEHIND THE PIPELINE
I like hard problems
with human stakes.
I combine a researcher’s skepticism with an engineer’s urgency. Whether I’m reconstructing a missing scene, designing safer conversational AI, or shipping a real-time product, the goal is the same: rigorous systems that create clear value.
Scouting leadership shaped how I work — stay calm under constraints, communicate clearly, and move a team toward the objective. I am now growing that discipline at NEXVIX as an AI Engineer Intern.
PYTHONPYTORCHC++JAVASCRIPTSQLGIT
08 START A CONVERSATION
Bring me the
difficult problem.
Hiring for an AI or software role? Building a product that needs vision, agents, or dependable engineering? Send the context directly.
Email me directly