PulseCheck

Inspiration

We were intrigued by T-Mobile's problem statement and the challenge that calculating the customer happiness index would be. PulseCheck links the customer happiness index to network coverage, arguably the most important factor to T-Mobile, and other factors such as internet speed and customer reviews from social media websites such as X.

What it does

PulseCheck is an app that connect T-Mobile with its heart: its customers. PulseCheck has a dashboard that displays the top 5 counties in the US, with the highest satisfaction scores, and the bottom 5 counties in the US, with the lowest satisfaction scores. PulseCheck also has a global map that displays US states and counties plus statistics such as average customer happiness, predicted forecast, general customer sentiments, and problems reported.

How we built it

PulseCheck uses Nemotron, an NVIDIA tool, to calculate the average customer happiness index and provide explanations for easy comprehension. We use Gemini to get the specific numerical analysis of the data and accurately forecast values, falling back to a standard linear regression if Gemini fails. The dashboard and globe map are designed with React, Tailwind CSS, and Shadcn/ui, and Mathbox. To use customer sentiments towards the happiness index, we used a web scraper on websites like X and Reddit, where people are likely to post about their experience or suggestions for T-Mobile in addition to general information like outages, and analyzed it to extract key phrases that reflect issues that T-Mobile has done well on or can improve.

Nvidia Dataflow: PulseCheck uses NVIDIA Nemotron to analyze customer data and generate actionable insights. The workflow:

Pulls customer metrics from the last two quarters (6 months).
Prompts Nemotron to analyze trends, summarize key drivers of happiness, and forecast the next quarter (3 months). Multiple prompt versions ensure a reliable response.
Passes Nemotron’s analysis to Gemini for precise numerical forecasting, with a linear regression fallback if needed.
Produces structured outputs combining explanations and forecasts for practical decision-making.

This demonstrates multi-step reasoning, workflow orchestration, and tool integration—using AI to analyze data and produce actionable forecasts, not just respond to prompts.

Challenges we ran into

As it was our first time working on such a complex project together, along with starting from multiple points concurrently, we ran into quite a lot of challenges! We encountered a lot of irrelevant datasets until we found datasets with county-level data that had the factors we needed. In addition, merge conflicts in Github and API issues with Gemini and Nemotron prevented us from being able to analyze the data we found and combine different areas of our project until they were fixed. X's API went down unexpectedly, making us scramble for another solution to get the customer feedback we needed.

Accomplishments!

We are extremely proud of creating PulseCheck, an app that tracks customer happiness over social media and network information and predicts changes that help T-Mobile anticipate and resolve obstacles to the satisfaction of its customers! We learned a lot about working together in a team, like making sure to work on separate Github branches! In addition, for most of us, it was our first time working with one or more of the technologies, and it was enlightening to learn to use new technologies on a time crunch!

What's next for PulseCheck

In the future, we plan to fine-tune the customer happiness index, adding analysis of factors such as demographics, loyalty, and plan usage and pricing. In addition, T-Mobile continues to provide service to its customers beyond just the US with international plans that are useful in trips to Europe, Asia, etc. Expanding PulseCheck will help T-Mobile track customer satisfaction even while they are overseas, helping T-Mobile remain the best wireless carrier in the US and beyond!