Md. Mutasim Billah Abu Noman Akanda

Senior Machine Learning Engineer · Data Scientist · AI Researcher

3+ years delivering production ML across computer vision, NLP, and LLMs/VLMs—document OCR at 10K+ docs/day, multimillion-site crawling, and measurable cost and accuracy wins (e.g. ~40% infra savings, 4× memory reduction on vision models, 92%+ valuation accuracy). 2+ years as a CS teaching assistant. Peer-reviewed work in ACM and Springer.

  • Document AI & OCR
  • LLMs, VLMs, RAG & voice
  • Web-scale data & ML systems
  • Teaching & mentoring
Email LinkedIn GitHub +880 1773 272649
Portrait of Md. Mutasim Billah Abu Noman Akanda
10K+docs/day OCR scale
2M+sites (ad detection)
$50K+annual compute saved
92%+valuation model accuracy
2peer-reviewed publications

Professional summary

Senior Machine Learning Engineer with 3+ years shipping production systems: enterprise OCR and layout analysis (including low-resource languages), large-scale crawling and detection pipelines, and LLM/VLM-powered products (RAG, voice, APIs). Focus on validation, cost–latency–accuracy tradeoffs, and maintainable services.

Research: document layout analysis and applied ML; peer-reviewed papers at ACM SE ’24 and Springer (LNNS). Teaching: 2+ years as undergraduate TA in CS (programming, OOP, AI, software engineering)—labs, assessment, and project mentoring.

Stack highlights: Python, PyTorch, FastAPI, Docker, PostgreSQL/pgvector, async crawling (Playwright, Crawl4AI), and common LLM inference tooling (vLLM, Ollama). English (professional) and Bengali (native).

Research interests

Document AI & vision

Layout analysis; printed and handwritten OCR; scene text; structure recovery from PDFs and images, including Bengali (Bangla) and other low-resource settings.

ML, LLMs & NLP

Deep models for vision and text; efficient inference and quantization (including VLMs); RAG, retrieval, and semantic search.

ML systems & evaluation

APIs and microservices for ML; large-scale crawling and data pipelines; validation and cost–latency–accuracy tradeoffs.

Work experience

  1. Senior Machine Learning Engineer

    Apurba Technologies Ltd. On-site · Dhaka, Bangladesh · Sep 2025 – present

    Project: Borno OCR — funded by the ICT Ministry of Bangladesh

    Leading enterprise deep learning OCR for Bengali text recognition, document layout analysis, and automated PDF/image → editable DOCX.

    • Architected dots.ocr backend integration for Bengali, scaling to 10,000+ documents daily.
    • Optimized Qwen2.5-VL with 4-bit quantization: ~4× memory reduction and ~40% lower infrastructure cost.
    • Proprietary bounding-box alignment: +25% conversion accuracy, −80% manual corrections.

  2. Data Scientist

    Pixalate Inc. Remote · California, USA · Mar 2025 – Aug 2025

    Project: Ad Intel Crawler — AI-powered ad network detection

    AI-driven large-scale web crawling, automated ad network identification, and real-time media analysis for advertising intelligence.

    • End-to-end detection system: 73% accuracy across 2M+ websites, 50K+ pages/day.
    • Deterministic pattern matching: reduced AI dependency, $50K+/year compute savings.
    • Scalable crawler orchestration: 1M+ requests/day with asynchronous processing.

  3. AI Engineer

    Green Pants Studio Remote · Texas, USA · Jun 2024 – May 2025

    Project: Vendidit — AI eCommerce intelligence

    Large-scale scraping, automated fair market valuation, and intelligent data mapping.

    • Vendidit Scraper: +40% speed, 100K+ products/day.
    • OLS regression for valuation: 92% accuracy, −60% pricing errors.
    • Dockerized microservices: −50% deploy time, +35% reliability.

  4. Machine Learning Engineer

    Apurba Technologies Ltd. On-site · Dhaka, Bangladesh · Mar 2023 – May 2024

    Project: Bengali OCR — ICT Ministry funded

    Deep learning OCR for Bengali, layout analysis, and scene text for government digitization.

    • YOLOv8 document layout: 85%+ accuracy on Bengali documents.
    • 8-bit quantization: 2× faster inference, 50% less memory.
    • Co-authored peer-reviewed papers at ACM SE ’24 and Springer (ICTCS / LNNS).

  5. Teaching Assistant, Computer Science & Engineering

    University of Liberal Arts Bangladesh (ULAB) On-site · Dhaka, Bangladesh · Feb 2021 – Jan 2023

    Supported core undergraduate CS courses for cohorts of 30+ students per offering.

    • Courses: Introduction to Programming, Object-Oriented Programming, Artificial Intelligence, Software Engineering.
    • Led lab sessions and tutorials; prepared lab materials; graded assignments and projects.
    • Office hours and one-to-one mentoring, including guidance on AI/ML course projects.

Technical skills

Programming & development

Python (expert), TypeScript, SQL, RESTful APIs, modular service architecture

Machine learning & AI

PyTorch, TensorFlow, scikit-learn, Sentence Transformers, Hugging Face, faster-whisper, vLLM, Ollama, LLM integration, deployment, OpenCV, EAST/CRNN handwritten OCR, QAT, DataParallel

Deep learning & NLP

LangChain, RAG, ChromaDB, transformers, GPT/Qwen workflows, semantic search, prompt engineering

Backend & data

FastAPI, WebSockets, SQLAlchemy, PostgreSQL, pgvector, Pydantic, rate limiting, HTTP microservices

Web automation & discovery

Playwright, Crawl4AI, browser-use, BeautifulSoup4, Scrapy, async crawling pipelines

Frontend & DevOps

Gradio, Next.js, React, Docker, Docker Compose, GitHub Actions, CI/CD, Pytest, Vitest

Data science & analytics

Pandas, NumPy, Matplotlib, Seaborn, MLflow, Weights & Biases, statistical analysis

Tools & collaboration

Git, GitHub, Jira, Notion, Slack, Agile, project management

Research & Technical Projects

Production-oriented systems aligned with current work: OCR, matching platforms, voice AI, and RAG.

SenseScan

Bengali handwritten document OCR

Full-page Bengali handwriting: EAST + LANMS detection, CRNN recognition with QAT-ready checkpoints, reading-order assembly. FastAPI endpoints (/plugin, /v1/ocr/handwritten), optional Gradio UI, DataParallel multi-GPU.

Python · PyTorch · FastAPI · OpenCV · LANMS · Gradio

GitHub →

GradConnectAI

Supervisor discovery & matching

End-to-end platform: CV signal extraction, AI-driven professor/opportunity discovery, ranked matches with evidence and outreach drafts.

FastAPI · Next.js · TypeScript · PostgreSQL · pgvector · Playwright · Docker

GitHub →

AI Interview Bot

Real-time voice interviewer

WebSocket PCM streaming, VAD/turn detection, faster-whisper STT, Ollama LLM, multi-backend TTS. Optional RAG (ChromaDB + sentence-transformers); separate interview-engine for ATS-style scoring.

FastAPI · WebSockets · Ollama · ChromaDB · FFmpeg

GitHub →

RAG local chat

Document Q&A

Streamlit app with RAG for document-based Q&A—high query accuracy and fast responses in a local setup.

Python · ChromaDB · LangChain · Ollama · Streamlit

GitHub →

Real-time parking tracking

YOLOv9 + tracking

Vehicle detection and tracking at 30 FPS with 92%+ detection accuracy.

PyTorch · OpenCV · Ultralytics · YOLOv9

GitHub →

Publications

Education

University of Liberal Arts Bangladesh (ULAB)

Dhaka, Bangladesh · Jun 2019 – Feb 2023

B.Sc. Computer Science and Engineering

  • Summa Cum Laude — top 1% of graduating class
  • CGPA: 3.96 / 4.00
  • Honors: Vice Chancellor’s Honors List; Dean’s List Scholarship (3×)
  • Coursework: Machine Learning, AI, Data Structures & Algorithms, Software Engineering, Computer Vision, NLP

Honors & awards