Arman

Portfolio / Aftabnagar, Dhaka

Md. AkibulHasan Arman.

Software Engineer

  • Product & full-stack applications
  • Backend systems
  • Applied machine learning
View selected work

Work in view

Real project evidence

  1. ProductFrameSignal
  2. SystemsSlateDesk
  3. Applied MLCCTV Violence Anomaly Detection

Selected work

  1. 01Academic workflow platform

    SlateDesk

    Role-specific workspaces connect to an ASP.NET Core backend where authorization, concurrency, and deadlines remain authoritative.

    SlateDesk admin workspace showing academic operations, people, classes, assignments, and submissions at a glance.
    Admin operations overview
    SlateDesk teacher assignment ledger showing a published assignment, submission count, and workflow controls.
    Teacher assignment workflow
    Source

  4. 04Computer-vision research

    CCTV Violence Anomaly Detection

    An offline comparison of classical and deep-learning approaches for binary violence classification.

    1. 01Input

      Balanced 5,000-image working subset

    2. 02Process

      Images prepared at 160 × 160 pixels

    3. 03Compare

      Eight classical and deep-learning approaches

    4. 04Output

      Normal or anomaly prediction with offline alert artifacts

    5. 05Evaluate

      Frozen 750-image test split

    Research pipeline from CCTV frame input through preprocessing, model comparison, anomaly detection, evaluation, and offline alert output.
    Offline research and evaluation pipeline
    MobileNetV2 transfer-learning confusion matrix on the documented 750-image frozen test split: 342 normal and 363 anomaly samples classified correctly, with 45 errors.
    Evaluation example: MobileNetV2 on the frozen test split

    Best documented comparison

    Multilayer perceptron

    94.93%

    Accuracy

    99.10%

    ROC-AUC

    Balanced 5,000-image working subset · frozen 750-image test split · offline experiment, not production CCTV performance.

    Source
View the complete work index

Capabilities / Evidence

What the work proves.

Each area below points to a project that can be inspected in detail.

  1. 01

    Application systems

    Applications that connect role-aware interfaces to API rules, authorization, and stored data.

  2. 02

    Product interfaces

    Responsive interfaces that help people browse, compare, and complete an action across screen sizes.

  3. 03

    Applied ML and computer vision

    Research that compares models and reports evaluation context for classification, anomaly detection, and segmentation.

  4. 04

    Data and reporting

    Warehouse design and Power BI reporting that turn cleaned retail data into an inspectable analytical model.

Graphics experiment / GLSL

A shader, explained plainly.

This custom GPU program distorts a procedural pattern, emphasizes its edges, and applies a color threshold. It is a graphics study—not a trained model or an ML prediction.

Custom vertex + fragment shaderMove a pointer to shift the transformation
A procedural image transformed in three visible steps. The animated version loads only on eligible desktop devices.
  1. 01
    Source pattern

    A generated monochrome surface enters the shader.

  2. 02
    Edge emphasis

    Distortion and contrast make local structure more visible.

  3. 03
    Color threshold

    Selected values resolve into the orange signal layer.

Looking for the real model and evaluation? Read the CCTV case study.

More work / 08

More work.

Supporting research, data, and application work—kept secondary to the four case studies above.

Featured studies

Research and data systems / 03 projects

  1. Grouped bar chart comparing four anomaly-detection models across accuracy, anomaly precision and recall, F1, false-positive rate, ROC-AUC, and average precision.
    05

    Face-Based Anomaly Detection

    A research comparison of supervised and anomaly-detection approaches using LFW and FaceNet embeddings.

    Source
  2. Power BI retail sales dashboard showing year and country filters, sales, invoice and customer indicators, a monthly trend, and product and country charts.
    06

    Retail Data Warehouse & BI

    A dimensional retail warehouse and Power BI reporting project built from the Online Retail II dataset.

    Source
  3. Scatter plot comparing three wall-crack segmentation models by crack-only mean IoU and no-crack false-positive rate.
    07

    Wall Crack Detection

    Multidataset wall-crack segmentation research comparing three encoder-based model families.

    Source

Additional systems

Web applications / 05 projects

  1. 08

    TravelEase public homepage with navigation, a curated England journeys introduction, and actions to explore trips or learn about the platform.

    TravelEase

    A role-based server-rendered travel platform covering discovery, booking, demo payment, itineraries, and support.

    Source

  2. 09

    RoleBoard dashboard with five demonstration user roles, role permission cards, and a post-management table.

    RoleBoard RBAC

    A post and comment management demonstration with role-aware server-side permission checks.

    Source

  3. 10

    PulseFlow public homepage with an emergency workflow introduction and a command-centre interface preview.

    PulseFlow

    A PHP emergency-hospital application organized around patient, nurse, doctor, and administration modules.

    Source

  4. 11

    FUB Energy Monitor dashboard showing floor and status filters, active monitoring state, and simulated building-energy metrics.

    EWU FUB Energy Monitor

    A Flask dashboard for schedule-aware simulated building-energy telemetry and monitoring controls.

    Source

  5. 12

    GitHub source view of the EduConsult Pro custom WordPress plugin registering service content types.

    EduConsult Pro

    A local WordPress/PHP project containing education-consultancy custom theme and plugin work.

    Source
Open work index

About

I work across what a person sees and the rules that make it dependable.

The common thread is practical engineering: clear interfaces, explicit system rules, and results that can be inspected in the source.

Based in
Aftabnagar, Dhaka
Read the resume

Contact

Send a useful note.

Share the role, project, or problem you have in mind. A little context makes the reply more useful.

Email is the active contact path

Write to Arman directly.

Tell me what you are hiring for or building, and include any useful links or constraints.

Write an email