HeLM - A Medical Discharge Platform

The company appears to be a two-person team building a medical discharge platform using LLMs and retrieval-augmented generation (RAG) to help providers create chatbots for patient education post-discharge. The platform is described as designed to simplify the handoff of care from hospital to home, reduce repetitive nurse work, and improve access to information for patients.

The project is self-reported and unverified — no revenue, customers or traction data are provided beyond what the authors describe. It was submitted to a hackathon, suggesting early-stage development.

The single most important open question is: What level of clinical validation has been achieved, and how does the platform handle safety and compliance in real-world use?

• The description states that HeLM is a platform for medical providers.
• It allows providers to build educational chatbots using document bases of protocols and patient information specific to their practice.
• Patients ask plain-language questions about discharge instructions, and receive document-grounded answers.
• The system uses retrieval-augmented generation (RAG) and LLM agents to search documents and generate responses.
• It is designed to help with routine topics like medications, wound care, activity restrictions, follow-up appointments, and warning signs.
• When an answer is not clearly supported by the documents, it can direct patients back to their care team or emergency services.

Inference: The platform appears to be a tool for generating chatbots that are grounded in clinical documentation, using LLMs and RAG. It is not a direct patient-facing app but a provider-facing platform for building chatbots, with safety features built-in.

• The tagline states: “A platform for medical providers that simplifies and reduces costs of post-discharge patient care and education.”
• The description claims the platform is designed to reduce the burden on nurses who spend hours responding to repetitive patient concerns.
• It positions itself as a tool to improve efficiency, reduce costs, and address equity issues in healthcare.
• The authors state that it emerged from real clinical experiences and was validated with doctors and nurses.
• They claim that the platform could be a step toward improving healthcare equity and quality, starting at the critical moment of post-care discharge.

Inference: The positioning is evolving from a technical hackathon project to a clinical tool for improving patient education and care handoffs. The claims are aspirational, not yet proven in real-world use.

• The description states that the platform is designed for medical providers, such as trauma surgeons, weight loss specialists, hematologists, and others.
• It is built to help providers create chatbots for patient education post-discharge.
• The authors mention working with clinicians in validation rounds.

Inference: The ICP appears to be healthcare institutions or clinical teams that manage post-discharge care. However, no specific customer segments or use cases beyond general specialties are detailed.

• No evidence of pricing, revenue, or business model is provided.
• The description does not mention any monetization strategy or customer acquisition approach.

Not evidenced

• Built with: fly.io, llama-index, openai, pinecone, python
• Uses retrieval-augmented generation (RAG) pipeline and LLM agents
• The system is described as having evolved from a basic prototype to one with more robust safety features
• Includes a front-end for providers to review patient responses and identify confusion
• The authors mention several rounds of prototyping with clinicians

Inference: The technical stack suggests a modern, LLM-driven platform. It has moved beyond MVP into a more structured and validated prototype, but no evidence of production deployment or scalability.

• The authors state that they have validated the tool with doctors and nurses, and are now preparing for testing in patients.
• They mention being gearing up for testing in real clinical environments.
• The platform was submitted to a hackathon, suggesting early-stage development.

Inference: There is no evidence of revenue, customers or live usage beyond internal validation. The project appears to be in a pre-testing phase, not yet in production.

• No mention of competitors or market analysis.
• The description does not reference existing tools or platforms for patient education or chatbots in healthcare.

Not evidenced

• The platform is described as early-stage, submitted to a hackathon, and not yet in production.
• Safety and compliance are cited as major challenges — the authors state they had to speak with hospital regulation departments.
• There is no evidence of how the platform handles protected health information (PHI) or regulatory compliance.
• The team size is only two people, which may limit execution capacity.
• No evidence of traction, revenue, or customer validation beyond internal testing.

Inference: The biggest risk is lack of real-world clinical validation, and potential regulatory and safety issues in handling PHI. The small team size raises concerns about scalability and execution.

1. What specific regulatory or compliance frameworks have you engaged with, and how are you ensuring PHI is handled safely?
2. How many clinical validation rounds have been completed, and what were the key learnings from those tests?
3. What is your plan for scaling beyond the current team size and prototype stage?
4. Have you identified any specific healthcare institutions or providers that are interested in piloting the platform?
5. What are the key technical challenges you've faced in productionizing this system, and how have you addressed them?