DEV Community: Tasnem Islam Prome

I Built an IT Help Desk Ticket System Using Python, Streamlit, and SQLite

Tasnem Islam Prome — Mon, 09 Nov 2026 12:49:37 +0000

I wanted to practise building a practical application based on everyday IT support workflows.

My IT Help Desk Ticket System is a Python portfolio project for managing support requests.

Features

The local application supports:

Creating and assigning tickets
Tracking open, in-progress, resolved, and closed requests
Searching and filtering tickets
Editing and deleting records
Dashboard charts and CSV exports
SQLite persistence

Public demo

The public Streamlit version is read-only by default and displays five fictional sample tickets.

This prevents visitors from modifying shared data.

Limitations

The project is a single-user demonstration. It does not yet include authentication, role-based permissions, audit logs, or production security controls.

Future improvements include SLA tracking, notifications, and user management.

GitHub repository: https://github.com/tasnem-tech/it-helpdesk-ticket-system

Building this project helped me explore the relationship between IT support processes, databases, and user interfaces.

Building SentinelGraph AI: An Explainable Cybersecurity Investigation Prototype

Tasnem Islam Prome — Mon, 09 Nov 2026 12:45:19 +0000

Security dashboards often show alerts without making it obvious why an event was flagged.

I started building SentinelGraph AI to explore more transparent security-event investigations.

Project design

The prototype is designed around:

Synthetic authentication events
Explainable detection rules
Risk scoring and severity levels
Interactive security dashboards
Relationships between users, devices, and IP addresses
CSV export for investigations

Technology

Python, Streamlit, Pandas, Plotly, NetworkX, and Docker.

Current development status

This project is still incomplete.

The Streamlit dashboard code is available, but the core sentinelgraph/ Python package and automated tests still need to be uploaded.

The animated preview is a concept illustration, not a recording of a working application.

Future work includes restoring the missing source code, adding tests, validating the application, and developing machine-learning detection capabilities.

GitHub repository: https://github.com/tasnem-tech/sentinelgraph-ai

I'm sharing this as a work-in-progress project and welcome constructive feedback.

Building a Production-Oriented MLOps Prototype with Python and FastAPI

Tasnem Islam Prome — Mon, 09 Nov 2026 12:44:17 +0000

Training a machine learning model is only one part of building a usable ML application.

I created Production MLOps Platform to practise connecting model training, evaluation, prediction APIs, and interactive applications.

Architecture

The project uses synthetic customer churn data and explores:

Reproducible dataset generation
Logistic Regression classification
Model evaluation and saved artifacts
FastAPI prediction endpoints
Streamlit dashboard integration
Pytest and Docker

Current limitations

This is an educational prototype, not a production-ready MLOps system.

The repository currently has a source-folder mismatch: Python files must be reorganised into the expected src/ structure before the documented application commands can work correctly.

Roadmap

I plan to verify the complete application, introduce CI/CD, explore MLflow, add model monitoring, and improve API security.

GitHub repository: https://github.com/tasnem-tech/production-mlops-platform

The project is helping me understand the engineering work required beyond model training.

I Built TicketTriage AI: An NLP-Based Support Ticket Classification System

Tasnem Islam Prome — Mon, 09 Nov 2026 12:39:06 +0000

Customer support teams spend significant time reading and routing incoming requests.

I developed TicketTriage AI to explore how natural language processing can assist this workflow.

What the system does

Classifies support tickets by category and urgency.
Uses TF-IDF and Logistic Regression.
Returns confidence scores and token-level explanations.
Provides a FastAPI prediction service.
Includes a Streamlit interface.
Flags low-confidence predictions for human review.

Engineering workflow

The repository documents synthetic-data generation, model training, evaluation, automated testing, Docker configuration, and GitHub Actions.

Macro F1 is used as a primary evaluation metric because performance across different ticket categories matters.

Limitations

The training dataset is synthetic. Real support conversations contain more variation, ambiguity, and changing language patterns.

The system is intended to assist support agents, not replace human decisions.

GitHub repository: https://github.com/tasnem-tech/ticket-triage-ai

I'm particularly interested in improving model confidence and practical human-in-the-loop workflows.

I Built CareShield AI to Explore Suspicious Healthcare Record Access

Tasnem Islam Prome — Mon, 09 Nov 2026 12:34:29 +0000

Not every unusual hospital login is a security incident.

Healthcare professionals may access patient records outside normal working hours because of emergencies, night shifts, or changing patient assignments.

CareShield AI is my research prototype exploring how contextual information could improve electronic health record access monitoring.

Features

Synthetic healthcare access-event generation
Contextual behavioural analysis
Random Forest classification baseline
Interactive Streamlit dashboard
Investigation queue and risk indicators
Chronological model evaluation

Technology

Python, Pandas, NumPy, Scikit-learn, Streamlit, and Plotly.

Important limitation

The project uses synthetic events and labels generated from simulated rules. Model performance therefore does not establish effectiveness against real hospital cyberattacks.

Future research includes independent threat scenarios, model calibration, and comparisons with transparent rule-based detection.

GitHub repository: https://github.com/tasnem-tech/careshield-ai

My goal is to explore explainable, context-aware cybersecurity rather than treating every unusual event as malicious.

Why 99.83% Accuracy Can Be Misleading: My FraudLens ML Project

Tasnem Islam Prome — Mon, 09 Nov 2026 12:33:07 +0000

Imagine building a fraud detection model that achieves 99.83% accuracy but fails to detect a single fraudulent transaction.

That's the problem I wanted to explore with FraudLens ML.

The project examines credit-card fraud detection using a benchmark dataset containing 284,807 transactions, including only 492 fraudulent transactions.

What I explored

Logistic Regression, Random Forest, and XGBoost
Class imbalance and random oversampling
Precision, recall, F1, PR-AUC, and MCC
Leakage-safe train/test splitting
Interactive Streamlit visualisation

Why it matters

Accuracy alone can hide poor minority-class performance. A fraud detector must be evaluated on how many fraud cases it catches and how many legitimate transactions it incorrectly flags.

The project also distinguishes published research results from independently generated experimental results.

Limitations

The dashboard presents research benchmarks and supports uploading experimental results. Published benchmark scores should not be mistaken for results reproduced by my implementation.

GitHub repository: https://github.com/tasnem-tech/fraudlens-ml

What evaluation metric do you prioritise for highly imbalanced classification problems?

I Built ResearchLens AI to Make Data Exploration and ML Benchmarking Easier

Tasnem Islam Prome — Mon, 09 Nov 2026 12:30:15 +0000

Before training a machine learning model, I believe it's important to understand the data.

That's why I developed ResearchLens AI, an interactive dashboard for exploring CSV datasets, identifying data-quality issues, and comparing baseline classification models.

What it does

Examines missing values, duplicates, and data types.
Creates interactive data visualisations.
Compares Logistic Regression and Random Forest.
Reports balanced accuracy, macro F1, and confusion matrices.
Fits preprocessing steps on training data to reduce leakage.

Technology

Python, Streamlit, Plotly, Pandas, and Scikit-learn.

Limitations and next steps

This is an educational benchmarking tool, not a fully automated scientific validation platform. Grouped validation, temporal validation, SHAP explainability, and automated testing remain future improvements.

The biggest lesson from this project is that reliable evaluation matters just as much as model selection.

GitHub repository: https://github.com/tasnem-tech/researchlens-ai

I'd love feedback from developers and data scientists working on similar tools.