Jump to:

AWS will provide US $100 in AWS Credit to individuals registered for the Hackathon. In order to receive AWS Credits, you must complete this request form.

  • You MUST be registered for the challenge in order to receive AWS credits
  • AWS reserves the right to only provide credits to participants who demonstrated a valid submission idea, in the sole discretion of the Sponsor.
  • Please allow 2-3 business days to receive AWS credits
  • Create a free AWS account or have an existing AWS account to obtain access to Amazon Comprehend Medical’s free tier.

    • The AWS Free Tier includes select services available for 12 months following your AWS sign-up date, as well as additional service offers that do not automatically expire at the end of your 12 month AWS Free Tier.

    • Note: AWS Credits can only be applied to AWS services. Additional charges incurred by the Entrant for the use of AWS products are the responsibility of the Entrant. Entrant is encouraged to monitor their usage of free trials as to not incur additional charges.
  • AWS credit requests received after October 11 at 4pm EST will NOT be guaranteed receipt.

After receiving the credit code, redeem it here https://aws.amazon.com/awscredits/. You can check the duration of your free trial here.



  • Datasets & Example Use Cases:
    • MTSamples: Transcribed medical reports contain important clinical information. However, they often contain sensitive personally identifiable information (PII) and can be difficult to share broadly. MTSamples (https://www.mtsamples.com/) is a collection of sample medical reports generated by transcriptionists for training machine learning models and other analytic needs. How can AI tools like Amazon Comprehend Medical improve the value of clinical notes to both patients and providers? For example, could an algorithm detect evidence of undiagnosed conditions such as deep vein thrombosis (https://www.cdc.gov/ncbddd/dvt/data.html) and suggest preventative steps? Or perhaps it could advise providers against using potentially offensive language (https://link.springer.com/article/10.1007/s11606-020-06432-7)? Use your imagination to make clinical notes an important part of the patient-provider relationship.
    • Q-Pain: Pain management is an important part of many clinical treatments. However, pain management decisions are highly subjective and sensitive to bias. This increases the risk that natural language processing (NLP) algorithms trained on clinical pain management data will themselves produce biased results. Researchers at Stanford University developed the Q-Pain (https://physionet.org/content/q-pain/1.0.0/) data set to help measure social bias in pain management. It includes 55 detailed question-answer pairs that allow easy substitution of multiple racial- and gender-specific identifiers. ML developers can use these vignettes to prompt text-generation algorithms like the GPT series and gauge the impact of the substitutions. How can AI tools like Amazon Comprehend Medical help provide more context to this bias analysis? For example, perhaps the bias is more extreme for different pain management scenarios or when certain types of entities appear in the prompt. This information can help ensure that medical NLP systems are as fair as possible for all patients.
    • OpenFDA: Post-market safety monitoring builds on data generated from clinical trials to ensure that drugs are safe for their intended use. Drug manufacturers must file reports of problems with prescription and over the counter (OTC) drugs with the FDA. The agency can then respond with additional requirements, such as changes to the packaging information. The OpenFDA (https://open.fda.gov/) project provides easy access to FDA data via a set of application programming interfaces (APIs). These data include adverse event reports, product labeling information, and recall enforcement reports. How can AI tools like Amazon Comprehend Medical help provide insight into drug safety? For example, perhaps certain manufacturers or active ingredients have a history of problems or product labels could be made more consistent? Use your skills to ensure that everyone can access and understand this data. Participants are free to use a single dataset or combine them to create their own version of a dataset they would like to use.