Data Analytics @ Ooredoo Maldives

I'm Mohamed Ziyan Zaid, a data analyst and Power BI specialist with 3+ years of experience. I also build deep learning and computer vision systems, and design how the results look.

See selected work Get in touch
Portrait of Mohamed Ziyan Zaid
Scroll
Years in data analytics
3+
Cancer classification accuracy
99.55%
Anomaly detection accuracy
93.8%
Inference per frame
4.73ms

01About

I sit where data, artificial intelligence and design meet, turning complex numbers into insight people can act on.

I'm a data analyst and Computer Science graduate (specialising in Data Analytics). Today I'm Assistant Manager for Data Analytics and Accounting at Ooredoo Maldives, co-founder and data analyst at Got To Go Maldives (a booking platform), and a part-time data analyst at HCK DEV since 2023.

My work covers data analytics, computer vision, deep learning and graphic design. I enjoy presenting results in a way that's clear and good to look at.

Education

  1. 2023 — 2025

    BSc (Hons) Computer Science

    Asia Pacific University, Malaysia · Data Analytics

  2. 2020 — 2022

    Diploma in ICT

    Asia Pacific University, Malaysia · Data Informatics · CGPA 3.71

  3. 2013 — 2015

    Secondary Education

    Dharumavantha School

02Experience

Where I've worked.

  1. — Present

    Assistant Manager, Data Analytics and Accounting

    Ooredoo Maldives · Full-time · On-site

    • Power BI
    • SQL
    • Python
    • Data visualisation
    • Data modelling
    • Forecasting
    • Automation

  2. — Present

    Co-founder & Data Analyst

    Got To Go Maldives · Booking platform · Maldives

    Co-founded a booking platform for the Maldives, where I also lead the data analysis.

  3. — Present

    Data Analyst

    HCK DEV · Part-time · Remote

    • Data analytics
    • Data modelling
    • AI modelling
    • Machine learning
    • Python
    • Graphic design

  4. —

    QA/QC Inspector

    MIFCO · Maldives

03Selected work

Research, built end to end.

01
  • Computer Vision
  • Deep Learning
  • PyQt6
  • YOLO

Multi-Model Deep Learning for Surveillance Anomaly Detection

An anomaly detection system for surveillance video that identifies and classifies 13 types of security threats, including theft, violence and vandalism, in near real time.

ImpactSupports SDG 16 (Peace, Justice and Strong Institutions) by improving public safety through automated threat detection.

Accuracy
93.8%
Inference
4.73ms
Throughput
25–30fps

Technical approach

  • Benchmarked five architectures: DenseNet121, EfficientGRU, DenseRNN, ViT-Temporal Transformer and CNN-LSTM with Attention.
  • Chose CNN-LSTM with Attention for the best balance of accuracy (93.80%) and inference time (4.73ms).
  • Integrated YOLO object detection for added situational awareness.
  • Built a multi-camera desktop app in PyQt6 for real-time monitoring.

Key features

  • Runs at 25–30 fps on standard hardware (NVIDIA RTX 4060).
  • Temporal smoothing to reduce false positives.
  • Two interfaces: live monitoring and an analytics dashboard.
  • Supports multiple concurrent camera feeds.
  • Intrusion detection mode with person-specific alerts.
02
  • Medical AI
  • Deep Learning
  • Healthcare

Histopathological Cancer Classification with Deep Learning

Deep learning models for automated cancer detection in tissue samples, distinguishing benign from malignant lung and colon tissue with 99.55% accuracy.

Clinical relevanceShows how AI-assisted pathology could cut diagnostic time and improve consistency, given the 10–30% disagreement rate among pathologists.

Accuracy
99.55%
Precision
98.8%
Tissue classes
5

Technical approach

  • Implemented two architectures: a custom CNN and a hybrid CNN-LSTM.
  • The CNN performed best, at 99.55% test accuracy.
  • The hybrid CNN-LSTM with attention reached 98.80% accuracy.
  • Tuned hyperparameters with a focused random search.

Key achievements

  • Perfect classification (100% precision and recall) for colon tissue.
  • 98%+ accuracy on the harder lung cancer subtype distinction.
  • Balanced performance across all five tissue classes.
  • Efficient processing of 768×768 px histopathological images.

04Toolkit

What I work with.

01

Data Analytics

  • Power BI expert
  • Python
  • SQL
  • R
  • Excel
  • Statistical analysis
  • Data visualisation
02

AI & Computer Vision

  • TensorFlow
  • PyTorch
  • Deep learning
  • Machine learning
  • Neural networks
  • Image processing
  • Object detection
03

Design & Video

  • Photoshop
  • Illustrator
  • Affinity
  • Premiere Pro
  • After Effects
  • DaVinci Resolve

05Contact

Let's build something worth looking at.