Historical applicant PDFs feeding a searchable admissions intelligence platform
Semantic themes tracked across accepted student personal statements
Application systems unified into one queryable applicant dataset
MedSchoolCoach
Boston, MA
Education Technology, Medical Admissions Consulting
Amazon Bedrock, Amazon Bedrock AgentCore, Amazon S3, AWS Lambda, AWS Step Functions, Amazon ECS (Fargate), Amazon DynamoDB, Amazon RDS
MedSchoolCoach’s physician advisors have long relied on experience and instinct to answer questions like which academic profiles get accepted where, or what themes separate a strong personal statement from a generic one. Avahi is building an AWS-native admissions intelligence platform that extracts and structures years of historical applicant data, tags personal statements and activities by semantic theme, and surfaces it through an analytics dashboard and an agentic conversational advisor built on Amazon Bedrock AgentCore, replacing manual pattern-spotting with data-backed guidance.
MedSchoolCoach is a Boston-based medical school admissions advisory firm providing personalized coaching, strategic guidance, and application support to pre-medical students across the United States. Physician advisors with admissions committee experience guide students through school selection, activity planning, personal statements, and interviews.
Advisors need to answer questions such as which experiences and personal statement themes are most common among students accepted to a given school, or how an applicant’s academic profile compares to that school’s published benchmarks, across 144 medical schools and three separate application systems (AMCAS, AACOMAS, TMDSAS), each with its own formats and essay prompts. That knowledge existed only as unstructured text buried in thousands of PDFs, making it impossible to answer these questions systematically or at scale.
MedSchoolCoach already operated within AWS, making it the natural platform for a solution spanning document processing, structured and semantic data storage, and generative AI. AWS gave the project scalable storage, serverless orchestration for extraction, and managed access to foundation models and agent infrastructure through Amazon Bedrock, all inside MedSchoolCoach’s own AWS account.
Avahi brought AWS technical depth together with applied AI expertise spanning document intelligence, data engineering, and generative AI. As an AWS Premier Tier Partner with hands-on Amazon Bedrock experience, Avahi was positioned to both architect and deliver a platform combining reliable data extraction with semantic understanding of unstructured essay content.
The platform starts with an orchestrated ingestion pipeline: an ECS Fargate task pulls historical application PDFs from Google Drive, AWS Step Functions coordinate metadata validation and extraction, and AWS Lambda functions run a multi-method extraction process, combining a structured-layout parser with Amazon Bedrock for complex or mixed-format documents. Extracted academic and activity data lands in Amazon RDS for structured queries, while Amazon DynamoDB tracks ingestion state across the pipeline.
On top of that foundation, Avahi is building a semantic layer: personal statements and application essays are converted into vector embeddings through Amazon Bedrock, enabling semantic search and classification against a taxonomy of more than 60 themes, from clinical exposure and research background to family circumstances and career narratives. That semantic layer powers a medical school admissions intelligence dashboard, giving advisors school-by-school academic and acceptance benchmarking alongside the most common personal statement themes and activity patterns among accepted students.
Advisors reach all of this through an agentic layer built on Amazon Bedrock AgentCore. A conversational advisor agent takes a natural-language question, retrieves the relevant structured and semantic data across the full applicant dataset, and returns a direct answer, giving advisors a chatbot interface to the platform instead of manually searching dashboards or PDFs.
Orchestrated ingestion pipeline (AWS Step Functions, ECS Fargate, AWS Lambda) processing 1,600+ historical AMCAS, AACOMAS, and TMDSAS applications
Multi-method extraction combining a structured-layout parser with Amazon Bedrock for GPA, MCAT, coursework, and essay content
Structured (Amazon RDS) and vector-embedding data stores supporting both exact and semantic queries
Semantic tagging pipeline classifying personal statements into 60+ themes and narrative archetypes
Medical school admissions intelligence dashboard with school-level academic and acceptance benchmarking
Agentic conversational advisor, built on Amazon Bedrock AgentCore, answering natural-language queries across the applicant dataset
Once complete, the platform gives MedSchoolCoach’s physician advisors a data foundation that spans 144 medical schools across three application systems, replacing manual PDF review and anecdotal pattern-matching with searchable, semantically-tagged insight into what actually drives admission at each school. Advisors will be able to ground every recommendation in the outcomes of thousands of past applicants rather than generic guidance.
MedSchoolCoach
Boston, MA
Education Technology, Medical Admissions Consulting
Amazon Bedrock, Amazon Bedrock AgentCore, Amazon S3, AWS Lambda, AWS Step Functions, Amazon ECS (Fargate), Amazon DynamoDB, Amazon RDS
MedSchoolCoach’s physician advisors have long relied on experience and instinct to answer questions like which academic profiles get accepted where, or what themes separate a strong personal statement from a generic one. Avahi is building an AWS-native admissions intelligence platform that extracts and structures years of historical applicant data, tags personal statements and activities by semantic theme, and surfaces it through an analytics dashboard and an agentic conversational advisor built on Amazon Bedrock AgentCore, replacing manual pattern-spotting with data-backed guidance.
MedSchoolCoach is a Boston-based medical school admissions advisory firm providing personalized coaching, strategic guidance, and application support to pre-medical students across the United States. Physician advisors with admissions committee experience guide students through school selection, activity planning, personal statements, and interviews.
Advisors need to answer questions such as which experiences and personal statement themes are most common among students accepted to a given school, or how an applicant’s academic profile compares to that school’s published benchmarks, across 144 medical schools and three separate application systems (AMCAS, AACOMAS, TMDSAS), each with its own formats and essay prompts. That knowledge existed only as unstructured text buried in thousands of PDFs, making it impossible to answer these questions systematically or at scale.
MedSchoolCoach already operated within AWS, making it the natural platform for a solution spanning document processing, structured and semantic data storage, and generative AI. AWS gave the project scalable storage, serverless orchestration for extraction, and managed access to foundation models and agent infrastructure through Amazon Bedrock, all inside MedSchoolCoach’s own AWS account.
Avahi brought AWS technical depth together with applied AI expertise spanning document intelligence, data engineering, and generative AI. As an AWS Premier Tier Partner with hands-on Amazon Bedrock experience, Avahi was positioned to both architect and deliver a platform combining reliable data extraction with semantic understanding of unstructured essay content.
The platform starts with an orchestrated ingestion pipeline: an ECS Fargate task pulls historical application PDFs from Google Drive, AWS Step Functions coordinate metadata validation and extraction, and AWS Lambda functions run a multi-method extraction process, combining a structured-layout parser with Amazon Bedrock for complex or mixed-format documents. Extracted academic and activity data lands in Amazon RDS for structured queries, while Amazon DynamoDB tracks ingestion state across the pipeline.
On top of that foundation, Avahi is building a semantic layer: personal statements and application essays are converted into vector embeddings through Amazon Bedrock, enabling semantic search and classification against a taxonomy of more than 60 themes, from clinical exposure and research background to family circumstances and career narratives. That semantic layer powers a medical school admissions intelligence dashboard, giving advisors school-by-school academic and acceptance benchmarking alongside the most common personal statement themes and activity patterns among accepted students.
Advisors reach all of this through an agentic layer built on Amazon Bedrock AgentCore. A conversational advisor agent takes a natural-language question, retrieves the relevant structured and semantic data across the full applicant dataset, and returns a direct answer, giving advisors a chatbot interface to the platform instead of manually searching dashboards or PDFs.
Orchestrated ingestion pipeline (AWS Step Functions, ECS Fargate, AWS Lambda) processing 1,600+ historical AMCAS, AACOMAS, and TMDSAS applications
Multi-method extraction combining a structured-layout parser with Amazon Bedrock for GPA, MCAT, coursework, and essay content
Structured (Amazon RDS) and vector-embedding data stores supporting both exact and semantic queries
Semantic tagging pipeline classifying personal statements into 60+ themes and narrative archetypes
Medical school admissions intelligence dashboard with school-level academic and acceptance benchmarking
Agentic conversational advisor, built on Amazon Bedrock AgentCore, answering natural-language queries across the applicant dataset
Once complete, the platform gives MedSchoolCoach’s physician advisors a data foundation that spans 144 medical schools across three application systems, replacing manual PDF review and anecdotal pattern-matching with searchable, semantically-tagged insight into what actually drives admission at each school. Advisors will be able to ground every recommendation in the outcomes of thousands of past applicants rather than generic guidance.
Let’s explore your high-impact AI opportunities together in a complimentary session
