Employee Attrition & Performance
Why do people leave? I built this dashboard to find out. Satisfaction fell the longer people stayed, even as their raises went up.

From the table, to the insight, to the model.
I build the databases people rely on, the dashboards that explain them, and the AI tools that run on top. Calm about the data, curious about the questions, and a little too fond of a good SQL pun.
Every number is a real dataset from my projects. Tap a card to see the work behind it.
By day I'm a Database Programmer at the Harris County Attorney's Office. I write the SQL behind case-management reporting and the Python automation that turns case data into finished legal documents, from first analysis through code review and UAT.
The chart beside this text is built from my Power BI attrition dashboard: nine job roles, bar height for headcount, and the top of each bar for the people who left.
My Master's in Data Science started with linear regression and ended with transformers. These days I use AI tools at work most days, and I build with them at home too.
Our team fine-tuned mBART and mT5 on 1.49 million English–Hindi sentence pairs. I handled the data prep, training setup and evaluation. mBART reached 98.7 F1 going from Hindi to English. mT5 kept tripping over idioms.
mBART · mT5 · Hugging Face · BERTScoreOur dropout model flags students who might leave, and SHAP shows the reasons behind each flag. A counselor shouldn't have to trust a score blindly. I worked on the data cleaning, the report and the model review.
Python · scikit-learn · SHAPI build websites, small automations and agents with AI assistants. This site is one of them, 3D panda and bad puns included.
Claude · ChatGPT · Three.js · PythonEach one links to its repo, with the notebook, the data and the full write-up.
Why do people leave? I built this dashboard to find out. Satisfaction fell the longer people stayed, even as their raises went up.
A class project I ended up leading, from cleaning the raw data to the final Tableau story across platforms, genres and publishers.
Spotting at-risk students early enough to help. I cleaned the data, wrote up the results and reviewed the model, and SHAP explains each prediction.
English ↔ Hindi with two transformer models. I built the training pipeline and ran the evaluation, and mBART beat mT5 by a wide margin.
Six decades of Japanese construction spending, five models, one forecast to 2030. Gradient boosting came out on top.
8+ years in tech, and every role taught me how data behaves in real life.
Where my career started, on enterprise systems and the data behind them.
Bigger enterprise projects, and the moment I chose data science.
A Master's in Data Science and research work, where these projects began.
Intern in 2025, now Database Programmer, running on SQL and Python.
A team where I can own the path from table to insight to model.