-
Chat With Search
-
Recent Reasearch Paper News Feed
-
Live Radio Stream of Newest Research
Inspiration
When people have health questions, they usually turn to Google—and that often leads to sources like Mayo Clinic or WebMD. At first glance, these seem like reliable places for medical advice, but the more we looked, the more we realized a major issue: they rarely cite actual research papers. Their articles can be vague, overly cautious, and sometimes even outdated. How can anyone be sure they’re getting the most accurate, up-to-date health information?
At the same time, AI can’t be blindly trusted either. While AI is great at summarizing information, without proper sources, it can hallucinate facts or misinterpret medical data. When it comes to health, misinformation isn’t just frustrating—it can lead to real consequences and bad decisions.
That’s why we built HealthByte. Instead of just generating answers, our AI finds, cites, and summarizes real, peer-reviewed research. Whether someone is looking for the latest treatments, nutrition science, or medical breakthroughs, we make sure they can see the actual studies, not just a simplified summary.
With HealthByte, users don’t have to wonder where their health information is coming from. They get AI-powered insights backed by real science, helping them make informed decisions with confidence.
What it does
HealthByte is an AI-powered platform that helps users find, summarize, and listen to open, public-access health-related research papers. Our AI scans vast medical databases, extracts key insights, and provides clear, concise summaries—so you don’t have to sift through dense academic texts.
When a user submits a health-related question, HealthByte's custom PubMed search engine finds relevant journal articles. The AI then stores these articles in a vector database, rewrites the query with precise medical terminology, and re-prompts itself to extract the most accurate, research-backed information. This ensures that users receive not just AI-generated responses, but fact-based answers sourced directly from peer-reviewed studies.
With our live radio feature, users can listen to real-time health news summaries completely hands-free, allowing them to stay informed without having to read through lengthy articles. Whether you're multitasking, commuting, or simply prefer audio content, our AI-powered radio ensures that the latest breakthroughs in medical research are always within reach, making cutting-edge health information more accessible than ever.
Behind the scenes, our backend continuously scans PubMed for the latest, most relevant peer-reviewed research articles. Once identified, our AI processes and extracts key insights, generating concise, easy-to-understand summaries in real time. These summaries are then converted into natural-sounding speech and seamlessly integrated into our live radio broadcast, ensuring users receive up-to-date, research-backed health insights without ever needing to read a single word.
To ensure true accessibility, HealthByte is designed with a high-contrast black-and-white color scheme, large, easy-to-read text, and a simple, intuitive layout. This it aimed towards making it easier for users with visual impairments or reading difficulties to navigate and consume information without distractions.
How we built it
HealthByte is built using LangGraph and LangChain for AI-driven search and summarization, powered by OpenAI’s models. We developed our own graph search toolchain for real-time access to health research and built a live radio experience.
The platform runs on a Next.js frontend with a Python (Poetry-managed) backend, all within a Docker-based infrastructure for scalability.
Challenges we ran into
We faced significant challenges in creating a seamless live radio station, spending hours troubleshooting Icecast and Liquidsoap. Issues ranged from OCaml package manager conflicts to audio dropouts and playback inconsistencies. In the end, we had to pivot to another system, and there’s still room for improvement.
Additionally, integrating LangChain for AI-powered search and summarization required us to build a custom toolchain to ensure fast and accurate responses when querying research papers. Managing Python package dependencies with Poetry in Docker also required fine-tuning for smooth deployment.
Accomplishments that we're proud of
- Successfully integrated AI-powered search and summarization while maintaining a clean, black-and-white interface for accessibility.
- Built a live radio stream, a unique feature that allows users to consume health information hands-free.
- Overcame technical and architectural challenges to develop a working platform.
What we learned
We learned everything from optimizing AI pipelines for fast summarization to handling real-time audio streaming and troubleshooting FFmpeg and Liquidsoap compatibility issues for seamless media playback.
Additionally, we gained deeper insights into accessibility needs, ensuring our platform remains inclusive and easy to use.
What's next for HealthByte
- Expanding beyond health research to cover physics, mathematics, computer science, and environmental science, making HealthByte a comprehensive research discovery platform.
- Adding multilingual support, enabling global users to access research in their native languages.
- Enhancing the live radio experience, as we were not able to get it running as smoothly as we had hoped within our development timeframe.
Log in or sign up for Devpost to join the conversation.