Devlog #2 - Pulse 🫀

What I got done

I started building the FastAPI gateway for model inference. I can now make API calls to predict mood. Also, I split the first Jupyter notebook into two for better individual model improvement.

What I’m struggling with

Working on the energy predictor model, I’m struggling to make it work. I think the problem might be my dataset because the input features should be correlated to the target.

Keeping you guys updated ✌️

Devlog #1 — Pulse 🫀

What is Pulse?

Pulse is an AI-powered daily health & habit coach. You do a quick check-in every day — sleep, mood, energy, stress, exercise — and over time Pulse learns your patterns, predicts how you’ll feel, and gives you a personal AI coach that actually knows your history. Think of it as a health journal that talks back.

What I got done

• ✅ Scaffolded the full project structure
• ✅ Sourced and explored Kaggle datasets for training
• ✅ Trained and exported the mood predictor model using RandomForestRegressor 🎉

Bumps along the way

Ran into a few pandas and scikit-learn issues that ate some time, nothing catastrophic, just the usual “why isn’t this working” moments that every ML project comes with. Got through them.

What’s next

• Train the energy predictor model
• Build out the FastAPI ML service with prediction endpoints
• Start on the NestJS backend — auth + check-in CRUD