Inspiration
Having a bad time with your car can be incredibly stressful. Whether it's having to miss shifts, less money in your pocket, or even not being able to afford your rent. We wanted to create a tool that could help people understand the domino effect of emergency scenarios to help them understand what they could do to stop the situation getting any worse. Using the idea of a chain of events we created BreakingPoint.
What it does
BreakingPoint helps provide insight into how an unknown cost may affect your financial situation over the next 30 days. By using the context of a bank, users can view an interactive impact graph and understand how repair costs, lost income, delayed income and future bills interact. Users can research possible solutions, understand requirements and see their new recovery plan with visible costs, assumptions and evidence to back up the calculations.
How we built it
We built BreakingPoint using Next.js, React, TypeScript, and Tailwind CSS for the frontend, Python and FastAPI for the backend, and React Flow for the interactive consequence graph.
Financial context with Nessie
Capital One’s Nessie sandbox API supplies our fictional customer’s accounts, transactions, deposits, and bills. This financial snapshot grounds each simulation in available cash, spending history, and upcoming obligations, showing why the timing of an emergency matters.
Simulating the crisis
Our deterministic engine models 30 days of shifts, paychecks, bills, and repair decisions. Estimated take-home income lost from missed shifts is:
$$ L=8rm\alpha $$
Here, \(r\) is hourly wage, \(m\) is missed shifts, and \(\alpha=0.80\) is the demo’s take-home estimate. Losses affect the paycheck covering those shifts, preserving the distinction between lost earnings and current cash.
The scheduled cash-flow projection follows:
$$ B_t=B_0+\sum_{d\leq t}I_d-\sum_{d\leq t}E_d $$
Here, \(B_0\) is starting cash, \(I_d\) is income received, and \(E_d\) is scheduled expenses. The peak funding gap is:
$$ S=\max\left(0,-\min_t B_t\right) $$
A separate constrained ledger tracks actual available cash and unpaid obligations, so later income cannot hide an earlier missed bill.
Finding interventions with SerpApi
SerpApi supplies web results, Maps listings and reviews, and transit directions to investigate transportation, repair providers, and assistance resources. Gemini guides research, while the backend validates evidence, limits API calls, and retains source links. Research priority follows:
$$ P=\text{Impact}\times\text{Urgency}\times\text{Actionability} $$
This focuses research on promising ways to interrupt the cascade. Listings never automatically establish affordability, eligibility, or personal approval.
Comparing recovery plans
The planner evaluates combinations of up to three actions from eight quantified candidates—at most:
$$ \binom{8}{1}+\binom{8}{2}+\binom{8}{3}=92 $$
Every combination reruns the simulator, accounting for action costs, preserved shifts, payroll timing, and overlapping effects. Plans are ranked using financial improvements, costs, feasibility, and unresolved requirements, with conditional outcomes clearly labeled.
Nessie grounds the scenario, SerpApi connects it to potential resources, and the simulation measures what each recovery strategy could change.
Challenges we ran into
The main challenges were ensuring that the data was correct when providing payroll dates and pending bills based on a number of factors. These factors included modelling the negative impact of an unaffordable repair leading to missed shifts and lower earnings in the future. Other challenges included incomplete or inaccurate data from providers. This includes the transport itinerary not respecting the desired journey time and ensuring more accurate validation on model responses. Lastly identifying potential solutions from solutions.
Accomplishments that we're proud of
We are proud of creating a solution linking together a finance dashboard, interpretable crisis scenario, referenced research and interactive recovery scenarios. Each recovery scenario has been calculated based on the finance timeline meaning the impacts can be clearly identified and linked back to the action taken. We have also made the uncertainty clear by having conditional requirements, supporting evidence and clear assumptions.
What we learned
We have learnt that when it comes to financial modelling it is important to take into consideration how much money you have at certain points in time and when bills fall due. We have also learnt that thorough validation is needed when using AI for research. Whilst searching results may show opportunities they do not provide evidence that it will work for an individual. Using AI to find opportunities alongside deterministic calculations helped us create an adaptable and transparent experience.
What's next for BreakingPoint
We would like to look into other scenarios outside of car breakdowns. These include reducing working hours, unexpected medical costs and disruption to housing. We would also like to improve the transport validation, save recovery sessions and allow more complex financial scenarios. Most importantly we want to trial BreakingPoint with users and financial support organisations to see if the explanations and next steps are helpful in a stressful situation.
Built With
- fastapi
- gemini
- nessie
- next.js
- python
- react
- react-flow
- serpapi
- tailwind
- typescript
