Md. Mutasim Billah Abu Noman Akanda

Senior Machine Learning Engineer | Data Scientist | AI Engineer


Profile picture of Md. Mutasim Billah Abu Noman Akanda

Research Interests

Education

University of Liberal Arts Bangladesh

Bachelor of Science - Computer Science and Engineering

Dhaka, Bangladesh

June 2019 - February 2023

CGPA: 3.96/4.00

Courses: Programming, Differential and Integral Calculus, Co-ordinate Geometry and Linear Algebra, Statistics and Probability, Digital Image Processing, Artificial Intelligence, and Introduction to Robotics.

Govt. Mohammadpur Model School and College

Higher Secondary School Certificate - Science

Dhaka, Bangladesh

February 2014 - February 2016

GPA: 5.00/5.00

Mohammadpur Government High School

Secondary School Certificate - Science

Dhaka, Bangladesh

January 2012 - January 2014

GPA: 5.00/5.00

Publications

1. Md. Mutasim Billah Abu Noman Akanda, Maruf Ahmed, AKM Shahariar Azad Rabby, and Fuad Rahman. 2024. Optimum Deep Learning Method for Document Layout Analysis in Low Resource Languages. In Proceedings of the 2024 ACM Southeast Conference (ACM SE ’24). Association for Computing Machinery, New York, NY, USA, 199–204. https://doi.org/10.1145/3603287.3651184

2. Akanda, M.B.A.N., Prodhan, M., Sarwar, S., Raatul, A.M., Paul, B. (2023). Voice Controlled Home Automation with Cloud-Based Environment Monitoring System. In: Joshi, A., Mahmud, M., Ragel, R.G. (eds) Information and Communication Technology for Competitive Strategies (ICTCS 2022). ICTCS 2022. Lecture Notes in Networks and Systems, vol 623. Springer, Singapore. https://doi.org/10.1007/978-981-19-9638-2_21

Work Experience

Apurba Technologies Ltd.

Senior Machine Learning Engineer

On-site (Dhaka, Bangladesh)

Sep 2025 - Present

Assigned Projects: Borno OCR - Government-funded Bengali OCR system for document digitization

Key Achievements: Architected dots.ocr backend processing 10,000+ documents daily with real-time processing capabilities. Optimized Qwen 2.5 Vision Model with 4-bit quantization, achieving 4x memory efficiency and reducing infrastructure costs by 40%. Developed proprietary bounding box algorithm, improving accuracy by 25% and reducing manual corrections by 80%.

Pixalate Inc.

Data Scientist

Remote (California, USA)

Mar 2025 - Aug 2025

Assigned Projects: Ad Intel Crawler - AI-powered ad network detection system

Key Achievements: Engineered ad network detection achieving 73% accuracy across 2M+ websites with real-time analysis. Implemented deterministic pattern matching algorithms, saving $50K+ annually in compute costs. Designed scalable crawler handling 1M+ requests daily with asynchronous processing.

Green Pants Studio

AI Engineer

Remote (Texas, United States)

June 2024 - May 2025

Assigned Projects: Vendidit - AI-powered eCommerce intelligence platform

Key Achievements: Developed Vendidit Scraper for eCommerce data extraction, increasing speed by 40% and processing 100K+ products daily. Trained OLS regression model achieving 92% accuracy for fair market valuation, reducing pricing errors by 60%. Implemented Dockerized microservices architecture, reducing deployment time by 50% and improving reliability by 35%.

Apurba Technologies Ltd.

Machine Learning Engineer

On-site (Dhaka, Bangladesh)

Mar 2023 - May 2024

Assigned Projects: Bengali OCR - Government-funded document digitization system

Key Achievements: Trained YOLOv8-based Document Layout Analysis model achieving 85%+ accuracy on Bengali documents. Optimized DLA model using 8-bit quantization, achieving 2x faster inference speed and 50% memory reduction. Co-authored research papers published in ACM SE'24 and Springer conferences.

University of Liberal Arts Bangladesh

Teaching Assistant, Department of Computer Science and Engineering

On-site (Dhaka, Bangladesh)

Oct 2022- Jan 2023

Assigned Courses: Introduction to Computer Programming, Structured Programming, Artificial Intelligence, and Software Engineering.

Responsibilities: Conducting tutorials and labs, reviewing sessions, helping with coursework evaluation, preparing course files, and counseling students.

University of Liberal Arts Bangladesh

Peer Mentor, Student Affairs Office (SAO)

On-site (Dhaka, Bangladesh)

Feb 2022- Nov 2022

No. of Students: 30

Responsibilities: Mentoring a group of students very closely both in academic and non-academic life, guiding students with proper resources.

University of Liberal Arts Bangladesh

Student Prefect, Department of Computer Science and Engineering

On-site (Dhaka, Bangladesh)

Feb 2021- Jan 2022

Assigned Courses: Structured Programming, and Object-oriented Programming.

Responsibilities: Preparing assignments, evaluating student works, counseling students, and preparing grade reports.

Honors & Awards

Community Activities

University of Liberal Arts Bangladesh

President, ULAB Computer Programming Club

On-site (Dhaka, Bangladesh)

Jul 2022- Feb 2023

Responsibilities: Communicating with the internal and external communities of ULAB for organizing various events, and delegating the required tasks to the other executive members.

University of Liberal Arts Bangladesh

Vice President, ULAB Computer Programming Club

On-site (Dhaka, Bangladesh)

Feb 2022- Jul 2022

Responsibilities: Assisting the President as required in the fulfillment of tasks and goals, and leading other executive members as required.

Skills Summary

Programming & Development: Python (Expert), RESTful APIs, Microservices Architecture, Git, GitHub, Version Control

Machine Learning & AI: PyTorch, TensorFlow, Scikit-learn, OpenCV, Hugging Face, YOLO, vLLM, MLOps, Model Deployment

Deep Learning & NLP: LangChain, RAG, ChromaDB, Ollama, Transformers, BERT, GPT, Large Language Models, Fine-tuning

Cloud & DevOps: AWS (S3, EC2, ECS, ECR, Lambda), GCP, Docker, FastAPI, CI/CD, Kubernetes

Data Science & Analytics: Pandas, NumPy, Matplotlib, Seaborn, MLflow, Weights & Biases, Statistical Analysis

Web Scraping & Automation: Playwright, Apify, Scrapy, BeautifulSoup4, Selenium, Data Pipeline Development

Tools & Collaboration: Jira, Notion, Slack, Agile Methodologies, Project Management, Team Leadership

Soft Skills: Research, Communication Skills, Team Leadership, Public Speaking, Analytical Skills, Presentation Skills, Time Management, Problem Solving

Achievements

Machine Learning Projects

RAG-based Local Chat Box

Streamlit application with RAG for document Q&A, achieving 90%+ accuracy. This project allows users to index documents, create embeddings, and interact with their data through an intuitive chat interface powered by state-of-the-art language models. Ideal for researchers and developers, facilitating efficient data retrieval and conversational interactions within a local environment.

Tech: Python, ChromaDB, LangChain, Ollama, PyPDF, Pandas, Streamlit

GitHub Link: https://github.com/noman024/rag-based-local-chat-box.git

LLM-powered Web Scraping

Intelligent scraping system using ScrapeGraphAI, handling 10K+ pages daily. It enables users to extract data from any website without understanding the HTML or page layout using customizable graph pipelines and prompts. The scraped data is then structured and saved into JSON, facilitating easy analysis and further processing.

Tech: Python, ScrapeGraphAI, LangChain, WebKit, Ollama, Playwright

GitHub Link: https://github.com/noman024/Langchain-based-Web-Scraping-with-ScrapeGraphAI.git

Real-time Parking Tracking

YOLOv9-based vehicle tracking achieving 92%+ accuracy at 30 FPS. Created a tracking system that implements centroid tracking with Euclidean distance and YOLOv9 object detection. It allows for the extraction of frames from an overhead CCTV video, performs object detection on each frame to detect vehicles, and tracks their centroids across frames.

Tech: Python, OpenCV, PyTorch, Ultralytics, YOLOv9, NumPy, JSON

GitHub Link: https://github.com/noman024/parking-spot-tracking.git

Schema-based Web Scraper

Customizable scraper with 99%+ uptime, processing 5K+ requests daily. Created a web scraping tool designed to extract data from any kind of web page. It provides users with the flexibility to define any number and structure of schema as long as the associated attributes are present in the webpage's HTML elements.

Tech: Streamlit, FastAPI, JSON, BeautifulSoup4, Python Request Module

GitHub Link: https://github.com/noman024/web-scraper.git

Multi-Modal Document Analysis

Advanced document processing system combining OCR, NLP, and computer vision for intelligent document understanding. This system integrates multiple AI technologies to provide comprehensive document analysis capabilities for various document types and formats.

Tech: PyTorch, OpenCV, Transformers, Tesseract, Python

GitHub Link: https://github.com/noman024/multimodal-document-analysis.git

Real-time Sentiment Analysis API

RESTful API for real-time sentiment analysis of social media posts and customer feedback. This API provides fast and accurate sentiment analysis capabilities for various text inputs, supporting real-time processing and scalable deployment.

Tech: FastAPI, BERT, Redis, Docker, Python

GitHub Link: https://github.com/noman024/sentiment-analysis-api.git

Game Addiction Analysis with Neural Network (Deep Learning)

Created a neural network from scratch that can predict whether a student can have a physical disorder (eyesight or hearing issue) due to game addiction. The model was trained on a new dataset which was prepared based on some survey question-answers collected from different private universities in Dhaka, Bangladesh. The model has achieved 80.70% of accuracy.

Architecture: Neural Network

Tech: Tensorflow, Keras, Scikit-learn, Seaborn, Matplotlib, Numpy, Pandas.

GitHub Link: https://github.com/noman024/game-addiction-analysis-neuralnet.git

News Classification with Naive Bayes Theorem (Machine Learning)

Applied Multinomial Naive Bayes classifier that can classify news between ham and spam based on Naive Bayes Theorem. The model was trained on a dataset that is already provided in Pandas. The model has achieved 98.85% of accuracy.

Architecture: Multinomial Naive Bayes classifier

Tech: Scikit-learn, Numpy, Pandas.

GitHub Link: https://github.com/noman024/news-classification-with-naive-bayes.git

Diabetic Classification with Support Vector Machine (Machine Learning)

Applied Support Vector Classifier that can classify diabetics between positive and negative based on some features collected from patients. The model was trained on a dataset collected from Kaggle. The model has achieved 80.51% of accuracy.

Architecture: Support Vector Machine

Tech: Scikit-learn, Seaborn, Matplotlib, Numpy, Pandas.

GitHub Link: https://github.com/noman024/diabetic-classification-with-svm.git