Project Sandman

• 

• 

• 

• 

• 

Inspiration

New parents face one of the hardest problems imaginable: sleep deprivation with no reliable data. Advice is inconsistent, pediatricians generalize, and every baby seems “different.”

As data-driven builders, we asked:

If we can model markets and weather, why can’t we model baby sleep?

So we built Project Sandman — a data science approach to one of the most universal, unsolved parenting challenges.

What it does

Project Sandman analyzes 18,621 real-world parenting stories from Reddit to uncover:

Which sleep-training methods actually work

How success varies by baby age

What parents tried before something finally worked

Instead of offering generic advice, we provide probability-based recommendations grounded in real outcomes.

Parents can input:

Age + Current Sleep Duration + Method Tried

And get:

A ranked probability of what is most likely to work next.

It turns emotional chaos into actionable, evidence-based insight.

How we built it

We built a full data pipeline:

Python + Reddit API to collect multi-year parenting data

Natural Language Processing (NLP) to classify sleep methods and outcomes

Statistical modeling to calculate success probabilities by age + method

Visualization to display patterns clearly

Challenges we ran into

Parents define “success” differently (4 hours vs 10 hours)

Many posts are emotional vents, not clear data

Babies go through sleep regressions, causing non-linear patterns

Extracting meaningful signals required extensive NLP tuning

We solved this by normalizing outcomes and analyzing trends relative to baby age.

Accomplishments that we're proud of

Built a dataset from thousands of real parents

Created a system that gives context-aware, probabilistic guidance

Delivered a tool that supports exhausted parents without judgment

Turned emotional stories into measurable, useful data

What we learned

There is no “one best method” — context matters

Methods change effectiveness as babies grow

Almost every success story follows multiple failures

Data-driven guidance is far more helpful than generic advice

What's next for Project Sandman

We plan to:

Build an interactive web interface

Allow parents to upload sleep logs

Improve prediction accuracy with real-time learning

Expand analysis to feeding schedules and naps

Ultimately, we want Sandman to become a data-powered parenting assistant.

Built With

• data-visualization
• hex
• hex.tech
• jupyter-notebook
• machine-learning
• matplotlib
• mining
• natural-language-processing-(nlp)
• numpy
• pandas
• praw
• python
• reddit
• scikit-learn
• text