
ai
AI-Powered Violence Detection System
Finalist in Innovation Challenge Season 13 – An intelligent real-time violence detection system using computer vision and deep learning
I build AI-driven products, payment integrations and enterprise-grade platforms. Then I teach other people how.
Currently integrating the NMI payment gateway for Hyjima's merchants, and building AI systems for businesses on the side: RAG pipelines, LLM features inside existing products, and automation that removes manual work. I run a product business of my own, and since early 2024 I have taught hands-on robotics and Python to international students at Venturas.

Now
Software Engineer · NMI
Jul 2025 – present
Not a list of every technology I have touched. These are the areas where I have shipped something real, for paying customers, and would happily do it again.
RAG pipelines over a company's own documents, LLM features wired into existing products, and agent-style automation for the work nobody wants to do by hand. Built to be reliable in operations, not just impressive in a demo.
Multi-tenant SaaS with role-based access, audit trails, backups, monitoring and the boring reliability work that keeps customers paying. I have founded and run a product business, so I build for margins, support load and uptime, not only for the launch.
Day job. NMI gateway integration for Hyjima, and before that Stripe, Google Maps, OpenAI and Firebase across client work. I care about idempotency, retries and what happens when the webhook never arrives.
MERN and Next.js web apps, Electron desktop software, and admin-heavy systems: a retail POS, an ERP, a hotel manager, an LMS. This site's routine generator and attendance dashboard run on Next.js and Supabase.
TensorFlow and OpenCV in production-ish settings: a real-time violence detector that reached the Innovation Challenge Season 13 final, a YOLO-based SmartCart checkout, and CNN image classifiers.
Core Web Vitals, structured data and Lighthouse work for client sites. This one ships Person and WebSite JSON-LD, a generated sitemap and sane robots rules, so it practises what it preaches.
Two overlapping tracks: engineering jobs during the day, and a teaching practice that started a year before Avian and never stopped.

NMI · Remote · US payments company
Integrating the NMI payment gateway for Hyjima's merchants: transaction flows, SDK work, and the reliability details that keep card payments boring.
Avian · Mirpur DOHS, Dhaka · on-site
Cross-platform desktop software in Electron and Node.js, an internal SDK, and Python pre-processing pipelines feeding it.

Venturas & StemonBD · Dhaka · international student cohorts
Hands-on robotics and Python for international students at Venturas, following a Japanese programming curriculum: build the robot first, then program it. Still running alongside the day job.
North South University · Dhaka
Coursework in deep learning, computer vision and neural networks. Coordinated the R&D team at the NSU ACM Student Chapter on the side.
One from the ML side, one desktop product, one web platform. Each has screenshots, the stack, and what I would do differently.

ai
Finalist in Innovation Challenge Season 13 – An intelligent real-time violence detection system using computer vision and deep learning

web
Complete retail POS and inventory management system with sales tracking, stock management, and reporting

web
Intelligent Customer Relationship Management system powered by AI for automated insights and customer engagement
Percent bars say nothing. These three tiers do: how recently and how deeply I have used each tool.
What most of my recent commits are in.
Shipped real features, would pick up again without a refresher.
Used on specific projects. Would need a day to get back up to speed.
T-0
Every cohort starts on the pad: syntax, logic, a first program that actually runs.
MAX-Q
Then real builds. Robots that move, APIs that answer, models that learn.
ORBIT
Then they leave: portfolios, competitions, university applications.
Since February 2024 I have taught hands-on robotics and Python to international students at Venturas, following a Japanese programming curriculum: build the robot, then make it think. Small cohorts, real hardware on the desk, and a working project at the end of every module.
Zero → Professional
Fundamentals, OOP, data pipelines, first ML models, and a production-style REST API to finish.
Build Real Systems
Electronics, Arduino, Raspberry Pi, sensors, and a small autonomous robot by the end. The track I run at Venturas.
Master Problem Solving
Big-O to dynamic programming and graphs, drilled through contest-style problem sets.
NSU ACM Student Chapter · 2022–23
Ran the research and development team, organising workshops and hackathons for the chapter.
NSU ACM Student Chapter · 2021
One-on-one academic mentoring for juniors, plus an introductory Python workshop for the global Hour of Code.
2019 – present
ICPC participant, 1st runner-up at the NSU ACM SC intra contest, 500+ problems solved.
06 Contact
I take on remote engineering roles, scoped contract work, and AI integration projects for businesses, and I run private cohorts on request. Email is fastest. I usually reply within a day.