Why did you choose to attend Hack the North?
a) I wanted to be in an environment where, for 36 hours, building product and writing code was the singular focus.
b) Hackers possess the skills to make beautiful things, break ugly things and defy the rules others have made (other hackers and non-hackers). Hackers care about the shipping velocity of code; they light up when they see a git merge request. Hackers deny there’s a problem they can’t solve; armed with an unbelievably deep understanding of software architectures, system administration, and programming languages, they figure things out. Hackers remain the most exciting people in the world.
c) Whiteboards that look like foggy windows, chairs that remind you of your childhood primary care office, and snack tables filled with food you’re allergic to can’t be found on Zoom.
d) The ratio of time I spend in VSCode versus Notion is a proxy for how quickly I’m growing. I spent much time in VSCode at HTN.
e) My best friends live in Toronto.
What did you build?
A personal health dashboard that uses wearable data (Apple Watch and 8Sleep bed), medical records (every health report since you began visiting the doctor), and environmental data (UV Index, AQI) to create an overall health score that increments via the completion of personally recommend tasks.
a) Overall health score: We used five metrics to sort the collected data (referenced above) into blood, nutrition, environment, exercise, and sleep. Each metric had a percentage score representing a user’s quantitative status in the five categories. We used 5 XGBoost decision trees to generate a feature score for each input type. For example, if your A1C levels were low at your last blood test but all other values were average, your score would be 93%. We weighted metrics according to their linkage to disease and other collected vitals.
b) Personally recommended tasks: Because we had 5 XG Boost models (one for each metric), we could analyze the gain for each model. In an XGBoost model, the payment is the most valuable attribute to interpret the value of each feature. The features with the highest income would result in the most significant increase in the score of 100. We wrote frontend code that displayed these recommendations, and, in the demo, you can view the health score incrementing by 2 when the blood report is read.
I built this project with Aahaan Maini (ML eng), Aditya Dewan (ML eng), and Leon Si (Full-stack). I designed the front-end and worked with Aahaan on pre-processing and finding datasets.
Whom would you like to thank for the experience?
a) Babson College Accelerator Fund for providing a stipend so I could enjoy this experience.
b) Pavan Jayasinha for the conversation at your lovely new apartment; Simran Mayra for the Friday night walk; Manroop Kalsi for the Saturday night walk; Mikael Haji, Kabeer Makkar, Aahaan Maini, Richa Pandya for a memorable silent disco party; Anush Mutyala, Dev Patel for setting the bar; Arnav Shah for always keeping the door open for a conversation and endless laughs; Sri Anumakonda, Heya Desai for checking in on everyone (one in-person, the other virtually); Aditya Dewan for teaching me about XGBoost; Leon Si for teaching me about JS frameworks; Aahaan Maini for canceling the $500 Nanonets API subscription so my mom wouldn’t get mad.
c) HTN staff. The experience is great when you don’t need to summon the staff more than three times. Will.i.am for the hilarious and thoughtful keynote (I tweeted questions at him).
d) Navid Nathoo for being transparent about the updates at TKS. I always enjoy talking to you.