EXPLORE THE WORK
Selected projects
Explore the result.
See my contribution, decisions, and lessons.
F/VApplications
Forge & Voice
I love books, but I’m often walking or busy with something else. Forge & Voice grew from wanting to enjoy those stories through audio.
Python · Fish Speech · AudioDocument import → reviewed script → cast recording. Includes per-speaker voice selection and recoverable exports.
Project story · Local studio walkthrough
WApplications
WiseShield
I wanted a way to check on the people close to me, especially older loved ones. WiseShield turns that concern into reminders and follow-ups for family, friends, and roommates.
Java · Spring Boot · NotificationsPersistent schedules, account-scoped inboxes, and missed-check-in escalation. Location sharing requires consent.
Project story · Local app walkthrough
±Machine learning
Sentiment Studio
Three networks can read the same review differently. Bringing them into one app showed why a model’s training vocabulary matters as much as its weights.
TensorFlow · CNN · BiLSTM · FNetThree saved neural networks, one review. Preserves each model’s training vocabulary; scores are not confidence estimates.
Try with sample text
◉Machine learning
Image Captioning
A fluent caption can still describe the wrong scene. This project connects image-to-text generation with a deliberate step to review and correct the result.
Computer vision · TransformerRestores the visual encoder, Transformer, and vocabulary together. Captions can be inaccurate and need human review.
Bring a photo
▥Machine learning
Housing Explorer
I wanted a prediction I could explain. Comparing models and preserving the evaluation led to an explorer with visible evidence behind its estimates.
scikit-learn · RegressionHeld-out R² 0.835 across 4,128 districts. Historical 1990 data; a reproducible extension of coursework.
Try with default inputs
SProject overviews
Socialdeck
Before players can connect, their accounts and profiles need to work. My contribution links those flows to the backend of our team’s application.
Flutter · Firebase · Team projectMy team contribution: backend integration for account, onboarding, and profile flows. Broader social features are in progress.
Project overview · no live demo
LMProject overviews
Language-model experiments
Training curves showed progress; generated text raised more questions. This coursework taught me to connect each conclusion to the experiment behind it.
BART · GPT-2 · TrainingHistorical coursework in fine-tuning and small-model training. Results are documented; runnable checkpoints are not yet recovered.
Research overview · no live demo