The AI Projects From Our B2B Hackathon

This episode of Out-Of-Pocket is brought to you by…

AI agents are rapidly becoming the next frontier in healthcare, but they also introduce cascading risk. When one model fails silently, the entire agentic workflow can produce harmful clinical outcomes.

We developed a practical guide that shows how to monitor AI agents end to end across predictive, generative, and agentic layers, including a real-world multimodal AI scribe use case.

Download the "Monitoring AI Agents in Enterprise Healthcare" guide: Download Here.

Our hackathon in NY, upping the production value

A couple months ago, we hosted our first AI Healthcare hackathon in New York with our cohosts Deerfield Management.

I mean the energy was crazy lol - we had carving stations, custom OOP mice, workshops on building AI gents, whiteboard marathons, and people hacking till 3 AM. I have to give huge props to the Deerfield team and their affiliate Cure’s space. They really went all out on the designs and branding + it’s a beautiful space. Please don’t expect that level of production from any future OOP event…

This time we wanted to have a hackathon focused on issues healthcare BUSINESSES deal with. It felt fitting as we brought the hackathon to NY that we actually started touching the profit & loss statement instead of just our dreams.

We had 60+ people, 13 teams, and a whole bunch of cool projects that were built from scratch in 36 hours. You can see them below with some notes from me. Reach out to the people that built these projects if you found them interesting!

And if you’re interested in coming to our next hackathon, you should sign up for the newsletter to stay in the loop. Reminder, we’re doing a hardware hackathon this spring in SF 👀. 

Hire Power - Finding Physicians To Recruit

[**Winner of People’s Choice 1st Place + the “I’d use this tomorrow” award]

Team Involved: 

• Bartosz Michalak - bartosz.michalak@themomentum.ai
• Gil Kazimirov - kazimirovg@gmail.com
• Janan Dave - janan.dave@verifiable.com
• Neelesh Mittal - neelesh.mittal@baysideops.com
• Saksham Arora - saksham.arora@savantbio.com

Summary: Hospitals today have more unfilled MD roles than ever, with each opening costing an estimated $200k/month in lost revenue and taking 4-6 months to fill. When provider roles sit open, patients wait longer for care, detour to the ER, and chronic conditions go unmanaged. Solving physician hiring quickly and effectively protects access, reduces spend, and ensures consistent clinical capacity.

Hire Power built a proprietary AI-enriched database that assesses both a physician's match score (licenses, specialty, training) and intent score (connections to providers at the health system, recency of graduation, social media activity, compensation analysis) to surface qualified candidates who are more likely to move. 

The solution automates personalized candidate outreach from recruiters, medical directors, or direct connections, enabling recruiters to close roles with relevant candidates fast.

Recorded Demo: Watch Here

Live Demo: Watch Here

[NK note: This project got an audible “oh damn” when they presented it. Anyone that’s run a healthcare business can tell you how hard it is to actually figure out which physicians to reach out to for recruiting. They’re not suddenly changing their LinkedIn to resume bullet points that give you the *hint hint wink wink* that they’re looking. 

This team had a pretty creative approach to a very real pain point, I’m not surprised that they won people’s choice]

Quick Interlude - Healthcare 101 Learning Summit!!!

Baited you with one project to get you interested, now I can shill things quickly.

Reminder, our healthcare 101 learning summit is 1/29-1/30 IN-PERSON IN NY! We’ll teach you how US healthcare works, you’ll meet other cool people in healthcare to build your network, and it’s gonna be a grand ole time. We have a limited number of seats for this because of the space, so if you’re interested sign up or email sales@outofpocket.health for a group rate.

And P.S. - If you want us to come teach your employees how to use AI more practically at their jobs, we run workshops here. Email sales@outofpocket.health and we can talk you through it.

VendlyAI: Assess your AI vendor's claims!

[**Winner - People’s Choice 2nd Place]

Team: 

• Josh Breite (josh.breite@gmail.com)
• Parth Shah (sparth99@gmail.com)
• Rebecca Reed (rebecca.reed11@gmail.com)
• Serena Wang (serenawang100@gmail.com)
• Sunnii Roh (sunnii.roh@gmail.com)

Summary

VendlyAI helps healthcare organizations cut through the noise of an overcrowded AI marketplace. Hospitals, health systems, and payors face thousands of competing vendors and spend significant time and resources evaluating them. VendlyAI simplifies this process by simulating vendor tools on an organization’s own data to identify which solutions deliver real value

For this hackathon, we built two proof-of-concept use cases: testing ICD-10 code extraction from clinical notes using Canvas Medical to simulate an EHR and PhenoML as one of the agents, and evaluating the success of voice scheduling AI using PhenoML. Behind the scenes, VendlyAI uses an LLM-as-a-judge to standardize these evaluations and generate clear, data-backed recommendations that help healthcare teams make confident, evidence-driven decisions about which AI tools to adopt.

Live Demo: Watch Here

[NK note: Another problem everyone is facing - I get an email every week asking which AI vendors are legit in X space. I think the idea of creating company specific sandboxes to test and just having vendors test their AI in there is clever.

It’s inevitable that we’re going to use AI to judge AI. Honestly, maybe our hackathon judges themselves should have been AI??]

Longitude - Streamlining the long-term care enrollment process

Team Involved:

• Jake Rigoglioso
• Aashna Shah
• Mehak Sharma
• Divyanshi Parashar
• Le Qi

Summary: Long-term care plans must complete detailed eligibility assessments for all applicants. Currently, this requires payer-contracted staff to spend hours calling patients, resulting in high admin costs, longer time to care enrollment, and inconsistent data quality.

Longitude’s VoiceAI-powered solution automates and streamlines the entire LTC eligibility process for payer staff — from initial patient screening to in-person assessments — using natural conversation, live translation services, and real-time structured data capture.

Live Demo: Watch Here

[NK note: This demo was a great use case in using AI to do the tedious calling tasks that many healthcare businesses are essentially mandated to do. The scarier side of this is that it kinda shows how many phone calls people that are cognitively impaired might get. Literally one part of the call was “do you live by yourself and have dementia?”. Like this could be a great business for long-term care plans, or home burglarers.]

DMEasy - Making durable medical equipment fulfillment easier

Team involved:

• Jamie Gong 
• Henry Chan
• Kailin Xu
• Sherry Chen
• Natasha Najdovski

‍Project: Inspired by a frustrating patient experience (ask Natasha about her dad's journey with diabetic insoles), our team examined the DME supplier workflow. Fulfilling a DME order is a multi-step process that often requires aggregating documentation from multiple providers to satisfy nuanced insurance requirements. To ease the coordination burden at the heart of DME fulfillment, we built DMEasy, an AI-powered workflow management system for DME suppliers. 

Once a prescription is received by a supplier, DMEasy automatically determines the tasks and documentation required to fulfill it based on the patient's payment method/insurance. It then uses multi-modal AI agents to coordinate the completion of these tasks, aggregating all required documents from the appropriate parties. 

The platform includes both a supplier dashboard for managing orders and a patient tracker (think Domino's pizza tracker for medical equipment) for visibility into where their order stands in the process. Built using Replit for the UI, ElevenLabs for automated voice calls to providers, and Keragon workflows for certain orchestration tasks.

Live Demo: Watch Here

[NK Note: There’s some micro private equity fund rolling up mom-and-pop DME suppliers absolutely salivating just looking at this project.]

Baby Come Back - bring back lost patients

Team: 

• Taylor Senecal
• Sabeeh Siddiqi
• Natalie Curnes
• Aditya Dalal

Summary:

Baby Come Back is an attrition analytics and patient re-engagement platform designed for health systems and private practices. It identifies patients who have missed appointments or stopped returning for visits and reaches out to them via text to uncover the reasons behind their disengagement. The system classifies responses into actionable and non-actionable categories, such as “moved,” “changed insurance,” or “poor experience” for the latter, and “cost concerns,” “fear,” or “forgetfulness” for the former, to help providers prioritize where meaningful intervention is possible.

For actionable cases, Baby Come Back deploys AI agents that use behavioral science and personalized communication to re-engage patients. These agents are tailored to the patient’s specific barrier: empathetic motivational messaging for those who are anxious about care, or cost-transparency and deductible-based education for those with financial concerns. By combining patient sentiment analysis, segmentation, and automated outreach, the system helps providers reduce no-shows, close care gaps, and improve long-term retention.

Live Demo: Watch Here

[NK Note: I would love a text messaging channel where I could absolutely unload on the patient experience part of my appointment. I don’t think the datacenters are ready to handle that onslaught though.

This definitely feels like a project the dental clinics would lose their mind over though.]

Continuum - Care manager communications on steroids

[**Winner of the People’s Choice Award 3rd Place]

Team involved:

• James Leonard (Linkedin)
• Phil Weiss(Linkedin)
• Michael Kronovet (Linkedin)
• Tanishq Kancharla( Linkedin)
• Ali Alagha (Linkedin)

Project: Continuum is a continuous learning platform designed to help care managers close care gaps and support patients through critical transitions in care. Too often, outreach feels impersonal or ineffective because care teams lack the right context — patients aren’t in front of them, and their tools don’t surface meaningful insights. Continuum changes that automatically identify high-risk individuals from claims data, generating rich, dynamic profiles that include social barriers, preferences, and engagement history.

With Continuum, outreach becomes personal and purposeful. Care managers can see what matters most to each patient — from health literacy and transportation challenges to hobbies and conversational cues — and use AI-assisted recommendations to craft tailored emails, texts, or calls that resonate. The system even automates background workflows, like arranging transportation or scheduling follow-ups, allowing case managers to spend more time connecting and less time coordinating.

Live Demo: Watch Here

Recorded Demo: Watch Here

[NK note: Lol they had a patient that was a Shark Tank fan, and AI generated a message from Mark Cuban to get their screening done. Misinformation, but for good!

I do think if we want people to come in for regular screenings, it should feel like someone actually looked through my charts and said “oh these are all the things YOU need, Nikhil”. Not a message that feels like a trigger because I just turned a certain age]

Featured Jobs - Who’s Hiring???

A few companies that are hiring right now!

Morgan Health (Health Care Analytics Engineer - New York City, NY, Boston, MA, or Washington, D.C.)

• Looking for a new challenge for your healthcare data warehouse design, dbt, and product design skills? Join Morgan Health's collaborative data team to build scalable, user-centric warehouse solutions that advance our mission to improve quality, affordability, and equity in employer insurance. Apply here.

Double Blind Bio (Product Designer - San Francisco, Hybrid)

• Help us turn clinical trials into clinical smiles by making complex workflows feel delightful. DBB builds software + AI agents for sites and sponsors to run trials faster and better. We’re hiring our first product designer. Apply here.

Deductible Detective - Apply cash pay transactions to deductibles

[**Winner of the “Duct Tape Dynasty” award for most creative workaround with AI]

Team Involved:

• Louise Tanski
• Pratyush Kapur
• Uzair Khan

Contact: louise@amerahealthsolutions.com 

Problem: More patients are buying drugs through cash-pay marketplaces like Cost Plus Drugs, and self-funded employers are purchasing six-figure therapies directly from manufacturers. Those transactions never hit the PBM or claims system, which means they’re invisible to plan analytics, stop-loss reporting, and care management. To make this even more painful, several states are requiring payers to apply out-of-pocket drug spending toward members’ deductibles, forcing plans to manually review and convert transactions into claims.

Solution: We built a data ingestion engine that turns these direct-pay transactions into claims, so employers and payers can process and analyze them just like any other pharmacy claim.

Live Demo: Watch Here

[NK Note: Not being able to track cash transactions is a very big issue with pharmacy claims right now and will probably increase next year as more people are uninsured.

Now the actual utility of a project like this totally depends on if health insurers care. But if more states pass laws where you can apply your cash transactions to your deductible, then patients actually do have an incentive to track these.]

Ditzel: Pathology Triage & Tracking

[**Winner of the “Damn Son” award for most mindblowing project]

Team Members:

• Shree Nadkarni, MD 
• Matt Schwartz
• Sonia Xu 
• Darshan Kalola, MD, MPH
• Revanth Korrapolu

Follow-up Contact: shree.nadkarni@mountsinai.org

Summary: Ditzel is a pathology triage and tracking solution that streamlines turnaround time (TAT) visibility and inpatient discharge efficiency. One of our team members' hospitals processes 200–300 surgical pathology cases daily, yet much of the workflow remains manual and siloed across PowerPath, email, and spreadsheets—creating delays that impact patient flow and hospital revenue. 

Our team built a machine learning–enabled dashboard that predicts pathology case completion times using synthetic, de-identified data from over 11,000 Mount Sinai cases. Ditzel offers an admin triage dashboard and clinician-facing tracker, with future phases including Epic/PowerPath integration and automated voice or chat notifications. Even a modest 5% reduction in pathology-dependent length of stay could yield over $4.7M in annual savings for a 1,000-bed hospital.

Live Demo: Watch Here

[NK note: I quite literally didn’t see this team for the entire 36 hours of the hackathon and then they emerged with their own fine-tuned model for pathology. Absolute maniacs, definitionally locked in.

Very cool to see a specialty specific workflow using AI - IMO this is where the power of the hackathon shines. Putting clinicians and engineers on the same team to come up with stuff like this.]

Undeniable - auto-appealing drug denials

Team members: 

• Prithvi Addepalli (LinkedIn)
• Ryan Brill (LinkedIn)
• Khi Chou (LinkedIn)
• Eric Pan (LinkedIn)
• John Jarvis (LinkedIn)

Project contact: johnljarvis@gmail.com

Summary: Undeniable automates the process of appealing drug coverage denials, saving time and administrative expense for health systems and providers. Undeniable automatically integrates with health systems’ electronic medical records and leverages the published medical literature, treatment guidelines, and FDA labels to draft patient-specific appeals letters.

Live Demo: Watch here

[NK note: Insurance company is gonna have some new rules. If a letter has any of the following, it gets denied:

• Em dash
• “It’s not _____, it’s _____”
• “Here’s why:”
• A bunch of numbers and bullet points. Hey wait…]

OpenHand for Employers

Team involved

• Rahul Sompuram
• Chethan Sarabu
• Kaustubh Bhal
• Jenna Rajani
• Deesha Kumbham

Contact info for followups: rahul@openhand.health

Summary: OpenHand for Employers helps self-funded employers cut healthcare costs by guiding employees toward more cost-effective choices for procedures, labs, and medications. Our platform integrates claims and transparency data to show real-time savings opportunities

and employer-level insights on realized and potential spend reductions across key care categories.At its core is a patient-facing AI sidekick that listens, learns, and provides personalized cost guidance at the moment of care. It helps employees compare care options, understand their benefits, and choose high-value sites of care.

An AI sidekick helping patients save and employers spend smarter.

Live Demo: Watch Here

[NK note: ClosedFist for employers is when they REALLY need you to choose a cheaper option.

Honestly I wish employees could split the savings with their employer for making more cost-effective decisions like this. Just add it to their paycheck! Tools like this are great, but they gotta get paired with actual incentives on the patient side which feels like the missing piece.

Also this team did some actual math on real pharmacy claims data to calculate savings and lowest price alternatives for different drugs.]

Conclusion, Parting Thoughts, and Sponsors

I love hackathons. IMO the best way to actually build a network and make friends is to BUILD things with other people. Hackathons are one of the coolest ways to do that. We had a bunch of first time hackathon people come to this one, and we hope to see you at our hardware hackathon in SF this spring. Sign up here to get notified.

Couldn’t have done this without our co-hosts at Deerfield, highly recommend you check out their affiliate Cure space in New York cause it’s sick.

Big thanks to our Tier 2+ sponsors. 

• Camber - Camber Health is rebuilding the financial backbone of healthcare. Our billing platform automates claims, reimbursements, and payments end-to-end; eliminating spreadsheets, manual follow-ups, and esoteric payer knowledge that drain
• clinics today. We want to build a world where the quality of care wins —not the best paperwork. 
• Arlo - What if your health insurance company was actually on your side? That's what we're building at Arlo. We're a health plan for small businesses designed to eliminate the financial fear and complexity that prevents people from getting care. Our simple, transparent approach has taken us from zero to tens of thousands of members in under a year! 
• Canvas - Whether you're vibe coding or building agentic automation, build your hackathon project with Canvas developer tools. Canvas is the most programmable EMR on earth so you can accelerate anything: notes, orders, scheduling, claims, and payment workflows. See what you can do with the Canvas Plugins SDK, custom API endpoints, hundreds of triggering events, custom UI components, and more writable FHIR API resources than any other certified system.

We make all of our sponsors participate either by presenting, giving us tools to use, or datasets to play around with. If that sounds like something you’d want to sponsor, hit us up.

Thinkboi out,

Nikhil aka. “Hack streets back, ALRIGHT!”

Twitter: ​@nikillinit​

IG: ​@outofpockethealth​

Other posts: ​outofpocket.health/posts​

{{sub-form}}

‎If you’re enjoying the newsletter, do me a solid and shoot this over to a friend or healthcare slack channel and tell them to sign up. The line between unemployment and founder of a startup is traction and whether your parents believe you have a job.