MedMatch: An AI to Connect You with Life-Changing Trials!

MedMatch finds clinical trials tailored to your symptoms, helping patients access life-changing opportunities with ease.

Comment

Inspiration

Billions remain excluded from life-saving trials — over 90% of studies occur in high-income countries, leaving 80% of the global population behind as of 2024, with 10.2 million deaths annually from treatable diseases due to lack of access (WHO Global Clinical Trials Report 2024). In urban areas, jargons stall participation; in rural regions, awareness is nonexistent. MedMatch AI Fusion was conceived to bridge both divides — restoring hope and health to the 4 billion people left behind. If more participated, treatments and drugs could be tested faster, reaching the market 2-3 years earlier, potentially saving 1.5 million lives yearly by accelerating approvals (based on IQVIA Global Trends in R&D 2025, estimating 15% faster development with broader participation).

What It Does

MedMatch is your revolutionary AI-powered lifeline, transforming how you connect with tailored clinical trials! It obliterates the overwhelm of endless options, delivering razor-sharp ranked matches, vibrant maps, and detailed email reports. This game-changer tackles the global trial access crisis, potentially saving up to 2 million lives annually by turbocharging participation and slashing drug approval times by 3-5 years (according to IQVIA 2025 trends). With its intuitive design, it’s the ultimate tool to empower billions worldwide!

Alignment with UN Sustainable Development Goals (SDGs)

  • SDG 3: Good Health and Well-Being — Accelerates access to innovative treatments, potentially saving 1.2 million lives yearly through faster approvals. (sdgs.un.org/goal3)
  • SDG 10: Reduced Inequalities — Empowers underserved billions with trial matches. (sdgs.un.org/goal10)
  • SDG 9: Industry, Innovation and Infrastructure — Leverages AI to transform healthcare equity. (sdgs.un.org/goal9)
  • SDG 17: Partnerships for the Goals — Invites global health collaborations. (sdgs.un.org/goal17)

How I Built It

  • Data Foundation (TiDB): TiDB’s vector database powers the core, querying trial embeddings to feed raw data (from ClinicalTrails.gov) into the pipeline for precise matching.

  • AI Intelligence (Kimi AI, PubMed, RxNorm, MeSH, OpenFDA): Kimi AI drives smart ranking, auto-calling PubMed for cutting-edge research (e.g., diabetes fatigue studies), RxNorm for drug insights (e.g., insulin mappings), MeSH for medical term links (e.g., neuropathy terms), and OpenFDA for safety checks (e.g., adverse event rates), enriching data dynamically.

  • Cloud Infrastructure (AWS): AWS, with Lambda and EC2, fuels email delivery and app-serving, ensuring seamless, scalable communication of trial reports globally.

  • Mapping Solutions (OpenStreetMap, Folium): OpenStreetMap enhances user experience with interactive maps, pinpointing trial locations, while Folium adds visual clarity to the frontend.

  • Interface & Tools (Streamlit, Python, Boto3): Streamlit crafts a vibrant, user-friendly frontend with live logs and styled outputs. Boto3 for invoking Lambda to email results.

  • AI Enhancements (OpenAI, Sentence Transformers): OpenAI facilitates Kimi AI’s API interactions and tool calling, while Sentence Transformers computes high-dimensional embeddings from symptom inputs, enhancing the solution’s precision and intelligent trial matching.

Built With

  • TiDB (vector database) – High-performance database for storing and querying trial embeddings with lightning speed, enabling precise matching. (TiDB Official Docs)[https://docs.pingcap.com/tidbcloud/tidb-cloud-intro/]
  • Kimi AI (LLM ranking) – Advanced language model for intelligent trial ranking, integrated via OpenAI API for dynamic insights. (Moonshot's Kimi AI)[https://platform.moonshot.ai/docs/introduction]
  • AWS (Lambda, EC2) – Cloud platform powering email delivery and app-serving with scalable, reliable infrastructure. (AWS Documentation)[https://docs.aws.amazon.com/]
  • OpenStreetMap (maps) – Open-source mapping service providing interactive trial location visuals for user clarity. (OpenStreetMap)[https://www.openstreetmap.org/help]
  • PubMed (research), RxNorm (drugs), MeSH (terms), OpenFDA (safety) – External APIs auto-called by Kimi to enrich data with research, drug info, medical terms, and safety stats. (NIH, FDA)
  • OpenAI (Kimi API Integration), Sentence Transformers (Embeddings) – OpenAI enables seamless Kimi API interactions for tool orchestration, while Sentence Transformers generates high-dimensional vector embeddings from symptom inputs, enhancing precision in trial matching. (OpenAI, Hugging Face)
  • Folium (maps), Boto3 (AWS SDK), Python (core), Streamlit (UI) – Libraries for mapping, AWS integration, core logic, and a vibrant user interface with live updates. (Folium, Python, Boto3[https://boto3.amazonaws.com/v1/documentation/api/latest/index.html])

Challenges I Faced

  • Integrating Tools with Kimi: Syncing Kimi’s tool calls with PubMed, RxNorm, MeSH, and OpenFDA hit API errors like 400 “tool not found,” testing my patience as I debugged mismatches.
  • Wrong Disease-Symptom Mapping: A tight similarity threshold in TiDB’s vector search hid matches, mapping symptoms like vitiligo to unrelated diseases, a struggle I tackled after seeing random results creep in.
  • SQL Snags: Early SQL query errors, like “not all arguments converted,” nearly derailed the project, requiring careful fixes to ensure accurate trial pulls.
  • Resilience Built: Each hurdle, from API glitches to layout tweaks, strengthened MedMatch to serve billions with a robust, user-friendly solution.

What I Learned

  • Global Relevance Mastery: Mastered TiDB’s speed, Kimi’s intelligent tool-calling, and AWS’s robust power, unlocking global trial access with user feedback shaping a joyful, impactful interface.
  • User-Centric Design: Learned that intuitive design and perseverance are key, prioritizing real-world usability over complexity to solve tangible healthcare challenges.
  • Impactful Innovation: Realized even early prototypes can drive change, inspiring me to refine MedMatch for billions with resilience and practical solutions.

Accomplishments I’m Proud Of

  • Fully Functional Demo: A tool that ranks trials with API insights, maps locations, and emails reports, ready to save millions.
  • Early Agentic AI Breakthrough: As a data engineer, I dove into the Agentic AI world, mastering TiDB and Kimi AI, building this impactful solution in record time.
  • Seamless Integration: Overcame tech hurdles to create a cohesive pipeline, blending TiDB, Kimi, and external APIs effectively.
  • Global Impact Design: Crafted a user-friendly interface that empowers patients worldwide with hope and accessibility.

What’s Next

  • Global Pilots: Launch trials with health organizations in urban hubs and rural areas, aiming to save 1.2 million lives annually by accelerating drug approvals.
  • Enhanced Features: Expand trial data with multilingual support and integrate blockchain for secure, transparent data sharing.
  • Scalable Innovation: Develop a robust platform to transition from demo to a widely accessible tool, measured by user engagement and lives impacted.
  • Continuous Improvement: Refine algorithms based on global feedback to maximize effectiveness and reach.

Share this project: Link to Devpost

Built With

Share this project:

Updates