My Role
Team
Emma Rapp, Product Manager
Alejandro Lobos, UX Designer
Tools
Figma, Confluence, Jira
Timeline & Status
8 weeks — interactive prototype, dashboards shipped, project cancelled
Overview
Rhino Scheduling is a mobile and desktop web app built for the NYC DOE that connects schools with nurse staffing agencies to find replacements for school nurses that call-out sick.
In NYC, nearly every school has a student with a medical condition that requires a nurse to be on-site for the school to operate. The district was losing dozens of school days due to lacking a centralized and fast way to find replacements.
During the course of the project, I led the UX and visual design direction and defined generative AI use cases.
Highlights
Harnessing LLMs to automate healthcare scheduling and reduce operational inefficiencies.
Context
Wait, school's cancelled?
When nurses inevitably call in sick, a series of slow phone calls and emails are made by schools to staffing agencies in a scramble to find a replacement. These inefficiencies cause closures across the entire city, affecting nearly 2.3 million students and 1800 public schools.
A Hard Promise to Keep
Problem Space
Help, digital transformation needed!
Employee management among school nurses is low-tech and virtually non-existent. Several challenges have prevented software solutions from being created in the past.
1
Complex user roles. Multiple users must communicate to make even a simple replacement request, including school-specific nurses, staffing nurses, staffing agents, and school administrators.
2
Decentralized data. Currently, data regarding nurse qualifications, health needs, contact info, and available shifts are scattered across different spreadsheets and platforms.
3
Reactive, not proactive. The manual text and email message chain used to fill available shifts means health coordinators aren't planning for future staffing gaps and protecting against closures.
Opportunities
Hedging against common staffing pitfalls.
If personel and health data are collected in a consistent manner, there are rich opportunities for predictive features that identify and address likely staffing risks.
1
LLMs for tailored data analysis. Consolidating information into a relational database and recording actions like call-outs and cancellations across schools and agencies can help generate targeted solutions.
2
Automation for 24/7 coverage. Defining recurring processes like posting shifts, sending notifications, and entering new users can expand beyond the limitations of human administrators.
Discovery Research
Popping the hood.
The previous team assigned to the project had since left the company and our client contact was unresponsive to communication; without access to known stakeholders, I relied on secondary research and previous user flows to understand the industry practice for nurse staffing.
Secondary Research
I found state legislation and school policy surrounding student health supervision and compiled a list of all applicable schools and districts.
I also researched staffing agencies serving schools in each borough to define required staffing nurse and agency data fields.
Data Modeling
I created sequence diagrams in order to understand how each user type interacted with one another and which actions they could perform.
To aid developers with database creation, I created a UML diagram to map each attribute and action to specific user types. (Abbreviated version shown below)
Solution
No sick days for Rhino.
Nurse Dashboard
I chose to create a progressive mobile web app (PWA) for nurses, as it allows them to call-out sick from wherever they are, thus increasing the timeliness of their call-outs.
Administrator Dashboard
For administrators, I opted for a desktop web app to allow admins across institutions to monitor call-outs from any browser. Tags and filtering make it simple to for search records.
Health Needs
I created a form that allows users to create nursing certifications and link them with cooresponding health needs. This ensures that all needs are covered by appropiate training and consistent terminology is used.
Permissions Settings
In order to protect sensitive personal and health information, I created a permissions system to limit sensitive data access to only relevent user roles. This also prevents accidental modifications made to profiles and call-outs.
AI Solution
Work smarter, not harder.
After shipping the core dashboard and call-out features, I realized that tracking call-outs and personel changes across the entire school system would still be difficult for admins. Thus, I leveraged OpenAI's API capabilities to ideate a high fidelity AI chat bot and integrated automation system. View the demo.
Freeform Insight Generation
Rhino AI provides suggested data analysis prompts. This guided method of prompting helps users create meaningful questions that relate to staffing efficiency.
Suggested Actions
Using user input, Rhino AI provides actionable steps to spur users to address future operational concerns like uncovered shifts, patterns in PTO utilization, and more.
Automation Chaining
Users can save common prompts to speed up their workflow, as well as combine them to create automations. This feature realizes the benefits of generative AI insights and reduces the need for repetitive tasks.
Multimodal Input
Unlike other automation tools which use rigid templates with limited triggers and actions, Rhino AI allows users to pick from suggested prompts or simply type out a custom condition.
Metrics
Keeping kids in the classroom.
While the main business function of Rhino Scheduling was to reduce missed nurse shifts and centralize the call-out workflow, ensuring students receive quality education without interruption was my ultimate definition of success.
1
% change in school closures. This metric would provide insights into how effectively Rhino Scheduling curbs uncovered shifts, therefore enabling schools to remain open for instruction.
2
Time spent by administrative users. This metric reveals how much time the automation and AI features reduce the workload of employees, thus returning time to reinvest into other health initiatives.
3
Recurring AI prompts. While not a quantitative metric, tracking the most common tasks and questions users ask Rhino AI can identify native automations to add, thus reducing API call usage costs.
Impact & Learnings
Pioneering the future of enterprise solutions.
At the end of my internship at Relentless Agency, Rhino Scheudling was cancelled due to budget cuts and shifting project priorities. Despite this, I was able to help ship the product's core features and create a blueprint for future work in generative-AI. These design patterns can be applied to scheduling, inventory management, and many other use-cases to improve business outcomes.
1
AI isn't just a buzzword. Though the term AI has become common-speak among tech startups, generative tools provide immeasureable potential for discovering new ways to optimize operations.
2
Internal tooling can save businesses a LOT of time and money: Small issues like staffing can benefit greatly from digital transformation, as they allow for automation of time-sensitive processes that may otherwise snowball into larger business deficits.
3
Smooth execution requires a technical mindset: Creating UML diagrams made it very convenient to ensure all necessary data types and fields were accounted for. This reduced friction during handoff with backend engineers creating our database.
