
NumPyGrad
A deep learning engine with an inspectable studio
A Python/NumPy learning engine with a studio for training boundaries, tracing predictions, comparing models, navigating a rover, and recognizing drawn digits.
- Python
- NumPy
- Streamlit
- Pytest
SingaporeSoftware + AI + analytics
I'm Htet. I build full-stack products, AI workflows, and data tools. From a video editor to a Power BI dashboard, I care about how things work and what they help people understand.
Things I've built and explored
Different problems.
The same curiosity.
Applications, creative workflows, and the systems behind them.

A deep learning engine with an inspectable studio
A Python/NumPy learning engine with a studio for training boundaries, tracing predictions, comparing models, navigating a rover, and recognizing drawn digits.

An AI music video studio with a creative copilot
Turns songs into editable music videos, with scene review, a browser-based timeline editor, and Shades Assistant for creative feedback and edits.

A circular-retail discovery game
An arcade-inspired way to discover surplus goods, with inventory-grounded matching and a mystery reveal.
An accessible, intergenerational social feed
An intergenerational social feed with multilingual speech tools, slang translation, and safer link handling.
From messy inputs to dashboards, predictions, and statistical findings.

Patient-demand forecasting and capacity planning
Turns patient-inflow forecasts into a planning tool: test demand growth, estimate staffing needs, and compare waiting-time pressure across urgency groups.

A complete customer data preparation pipeline
Brings four regional spreadsheets through profiling, cleaning, identifier masking, and feature engineering into a structured Excel master table and customer analysis.

From audited workout data to evaluated calorie estimates
Audits repeated workout records, engineers relative heart-rate intensity, and diagnoses four regression models to explain the quality and limits of calorie predictions.

Survey research into student belonging and FoMO
A 16-item, 67-response study linking general FoMO and in-person versus online exclusion to campus wellbeing recommendations. My work covered quantitative calculations, interpretation, and findings synthesis.
A little about me
Based in Singapore
I like moving between the part you interact with and the machinery that makes it possible.
That has taken me from an accessible social feed and a music video editor to a gradient engine, hospital dashboards, and predictive modeling. I enjoy working through both the software and the data.
My background is in Applied AI & Analytics at Nanyang Polytechnic. Across projects, I try to make the decisions visible: what I built, why I chose it, and what I'd improve.
More about me