Skip to content
View harshadulshan's full-sized avatar
🐑
God is Good
🐑
God is Good

Block or report harshadulshan

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
harshadulshan/README.md
Harsha Dulshan: Learn Today. Build Tomorrow. Data Analyst, Maths Educator, BSc MIS at NSBM, AWS Certified Cloud Practitioner Data Analyst · Data Scientist · Educator. I turn messy data into decisions.

LinkedIn: harsha-kaldera Kaggle: harshakaldera Medium: @harshakaldera540 ORCID: 0009-0009-5824-0051
Facebook: harsha.dulshan.14 Instagram: @harshadulshankaldera Email: hphdkaldera@gmail.com

Profile views

👋 About me

I'm a data analyst who likes the whole journey: cleaning messy records, finding the real question, building the model, and explaining the answer so people can act on it.

whoami: Harsha Dulshan, Data Analyst, Data Scientist and Educator, Sri Lanka. BSc (Hons) MIS (Special), NSBM Green University 2022-2026; Data Analyst Intern, S & S Power Solutions Lanka 2025; 5+ years teaching O/L Maths & Science; researching which AI wrote a text; exploring image forensics with SAM 3, data engineering and quant finance; goal: university professor.

12 open-source data projects · 5+ years teaching Maths & Science · 14 certificates · 1,196 AI-written texts in my thesis

🚀 Featured projects

Click a card to open the repository. The numbers come from each project's own tests and benchmarks.

Which AI Wrote This? Macro F1 0.71 across 4 models, 2.8x chance Electricity Theft Detector: 88% of the top 500 flagged were confirmed thieves Code Plagiarism Detector: AUROC 0.86 on IR-Plag, JPlag 0.56, Dolos 0.72 Market Risk & Volatility: 24 VaR breaches vs 25 expected, Kupiec p = 0.84 Literature Review Screener: 95% of relevant papers found after reading 40% Exam Timetabler: 78 of 78 runs clash-free, median 0.9% from best known

🧪 6 more projects: exam questions, inventory, outbreaks, climate, earthquakes, air quality

Exam Question Analytics: finds the wrong answer key in SAT12 Demand & Inventory Optimizer: 99.7% fill rate at 15% lower total cost Outbreak Signal Detection: EARS C1-C3 and CUSUM alarms Climate Trend Analyzer: Colombo warming +0.10 °C per decade, p = 0.032 Aftershock Risk Forecaster: 34% chance of an M4.5+ aftershock within 7 days Air Quality & Health Risk: none of 8 big cities had good air on 29 Oct 2026

🧭 My journey

5+ years teaching O/L Maths & Science · 2022 BSc (Hons) MIS at NSBM · 2025 Data Analyst Intern · 2026 thesis and 12 projects · next: university professor

💼 Experience details

📊 Data Analyst Intern · S & S Power Solutions Lanka (Pvt) Ltd · Jul – Dec 2025

  • Audited data and resolved duplicate and inconsistent records; built a centralized customer database
  • Sales analysis, customer segmentation and cross-selling; inventory monitoring and demand forecasting
  • Dashboards and reports in Excel and Power BI

👨‍🏫 Mathematics & Science Teacher (G.C.E. O/L) · 5+ years

  • English and Sinhala medium at Zenith Educational Institute (Piliyandala), Sylvester Institute (Mount Lavinia) and Aruna Institute (Halpita), after starting with individual tuition
  • Turning exam results into insight led to the Exam Question Analytics project

🏗️ Kaldera Construction

  • Manage a construction business; built a digital workflow for project timelines, resources and financial reporting (SQL Server, Excel)

🛠️ Toolbox

Python, R, SQL, Java, C#; pandas, NumPy, SciPy, scikit-learn, XGBoost, SHAP; Power BI, Excel, Plotly, Streamlit, Jupyter; SQL Server, AWS, Git, GitHub, Numba

✍️ Research & writing

  • 🎓 Thesis: stylometric attribution of LLM-generated text under rewriting pressure. 1,196 samples from 4 models across 4 domains, rewritten lightly and heavily; an XGBoost + Random Forest + MLP ensemble with SHAP. Journal submission planned for November 2026.
  • 🖼️ Now exploring on Kaggle: recovering AI-edited images with Segment Anything Model 3 (SAM 3), plus data engineering and quantitative finance.
  • 🆔 ORCID: 0009-0009-5824-0051

📝 Latest on Medium

- [He Said He Was “Just Not a Maths Person”](https://medium.com/@harshakaldera540/he-said-he-was-just-not-a-maths-person-bb7aabbe318a?source=rss-4b28d0eb930e------2) · Nov 1, 2026

📜 Certifications

14 certificates from Google, Microsoft, IBM, AWS, Duke University and HP LIFE · click to see and verify
Issuer Certificate
Google Foundations of Data Science · The Power of Statistics · Go Beyond the Numbers · Share Data Through the Art of Visualization
Microsoft Preparing Data for Analysis with Microsoft Excel · Harnessing the Power of Data with Power BI
IBM Generative AI: Elevate Your Data Science Career · Generative AI Essentials for Data Science (Credly)
AWS AWS Cloud Practitioner Essentials · Exam Prep: AWS Certified Solutions Architect – Associate
Duke University Introduction to Probability and Data with R
HP LIFE Data Science & Analytics · Presenting Data · AI for Business Professionals

All earned in 2026.

📊 GitHub activity

3D contribution graph with stars, forks and languages Plasma reactor core: total contributions, current streak and longest streak Plasma serpent absorbing the contribution grid

💬 Ask me about data analytics & forecasting, turning exam results into insight, or research on AI-written text

Learn Today. Build Tomorrow.

Popular repositories Loading

  1. CarService-FlutterApp CarService-FlutterApp Public

    Dart 7 7

  2. harshadulshan harshadulshan Public

    Python 7 2

  3. VirtualPainter- VirtualPainter- Public

    AI Virtual Painter using Python, OpenCV and MediaPipe

    Python 6

  4. SmartAttendance SmartAttendance Public

    🔐 AI-Powered Smart Attendance System with Face Recognition, Live Dashboard & Auto Email Reports

    Python 6

  5. NSBM-Smart-Attendance-Destop-Application- NSBM-Smart-Attendance-Destop-Application- Public

    AI-Powered Smart Attendance Desktop System for NSBM Green University

    Python 6 1

  6. ResearchIQ ResearchIQ Public

    JavaScript 6