POC project to automate processing of claim documents to reduce manual effort and improve consistency
Claim documents were generated using the MAIB narrative dataset and uploaded to s3. Using a foundation model from AWS Bedrock, information can be extracted and a summary created.
The diagram below shows the architectural diagram for processing the documents.
Userflow
Step 1: get upload URL
GET /upload-url?filename=my_doc.pdf
Authorization: Bearer <JWT>
Step 2: upload directly to S3
PUT <uploadUrl>
Content-Type: application/octet-stream
Step 3: get summary
GET /summary?filename=my_doc.pdf
Authorization: Bearer <JWT>
Init terraform
Replace $ENV with your current env e.g. dev
terraform init -backend-config="key=envs/$ENV/terraform.tfstate"
Create terraform plan, then deploy infra
terraform plan -out "tfplan"
terraform apply "tfplan"
Install poetry
curl -sSL https://install.python-poetry.org | python3 -
pip install poetry-plugin-export python-inspector