For digital health startups like Mariana.AI, generating accurate and scalable AI clinical notes is mission-critical. To push the boundaries of compliance, cost-efficiency, and output quality, Mariana.AI partnered with GoML to explore Claude models via AWS Bedrock, benchmarking them against their existing OpenAI-based system. This collaboration led to a modular and future-ready architecture for AI in clinical notes, powered by Langchain, Portkey, and Claude Sonnet models, without disrupting existing workflows.
Mariana.AI is a cutting-edge healthcare innovation company focused on automating clinical documentation and patient engagement. To modernize its systems, Mariana.AI partnered with goML to migrate its clinical note generation and medical coding workflows from OpenAI to AWS Bedrock. The goal was to enhance performance, improve prompt engineering, and establish a scalable, compliant backend for future expansion into real-time and voice-based clinical support.
The problem: outdated AI for clinical notes that impacted accuracy
Mariana.AI had already deployed an OpenAI-based system that generated AI clinical notes and medical codes from transcribed doctor-patient conversations. While functional, the system faced limitations in terms of accuracy, tone, and clinical relevance. Additionally, the team wanted to reduce their dependency on a single vendor and explore alternatives that could offer better flexibility and long-term scalability.
Maintaining compatibility with their existing workflows was essential, so any evaluation needed to be non-disruptive. To address these goals, Mariana.AI partnered with GoML to benchmark Claude models, via AWS Bedrock, against their current setup, aiming to understand the performance trade-offs and potential improvements in AI clinical notes generation.
The solution: a modernized AWS stack for AI in clinical notes
GoML worked with Mariana.AI to build a standalone proof of concept that benchmarked Claude models (Sonnet 3.5, 3.7, and 4) against their existing OpenAI-based system. The goal was to evaluate clinical documentation quality across different model families, focusing on tone, completeness, accuracy, and clinical alignment. This setup was intentionally kept separate from Mariana.AI’s production backend, allowing for a clean comparison without impacting live workflows.
“This migration wasn’t just about switching APIs, it was about engineering trust, speed, and structure into every AI-generated report,” said Prashanna Rao, Head of Engineering at GoML.
AWS Bedrock and Langchain at the core of clinical AI orchestration
The modern AI in clinical documentation pipeline uses:
- Prompt Orchestration: Powered by Portkey and Langchain, routing structured prompts to Claude Sonnet Family Models.
- Claude Sonnet: Core LLM behind intelligent clinical report generation and medical coding.
Enterprise-grade architecture for scalable AI in clinical notes
- Preprocessing Pipelines: ECR containers extract and standardize notes from physician tools.
- Vector DB Indexing: Embeddings stored in OpenSearch for context-rich retrieval.
- Explainability Framework: A dedicated Chief Medical Officer (CMO) agent signs off on validated reports.
- CI/CD: GitHub, Docker, and CloudFront pipelines for fast iteration and DevOps readiness.
The impact: accurate, validated, and fast AI for clinical notes
Mariana.AI now benefits from consistent, high-quality clinical reports generated by AI, with explainability, benchmarking, and compliance built-in.
- 82% reduction in manual report verification time due to structured JSON validation and explainability checks.
- 3x faster faster generation of AI clinical notes over legacy systems
- 97% adherence to output schema across specialties like cardiology, neurology, and gastroenterology.
- 65% improvement in response accuracy after benchmarking with real-world inputs using F-score analysis.
“GoML’s solution gave us structured, accurate clinical notes with real-time validation. We’re now future-ready for EHR integration and voice-based reporting,” said a senior product manager from Mariana.AI.
Lessons from Mariana.Ai’s success with AI in clinical documentation
- AI in clinical documentation isn’t just about generating text, it’s about designing intelligent systems that ensure accuracy, compliance, and scalability.
- AI agents, when combined with schema validation and benchmarking, unlock real value in documentation workflows.
- Automated prompt orchestration and specialty-focused agents reduce dependency on manual edits and clinical QA cycles.
Advice for digital health teams
- Identify repetitive clinical documentation tasks, those are ideal for AI agent automation.
- Use multi-agent frameworks to separate logic, validation, and explainability for better control.
- Combine RAG + validation layers to boost accuracy and trust in AI-generated outputs.
- Ensure every Gen AI output is tied to a schema, structure is critical in clinical environments.
Ready to elevate your AI in clinical notes stack?
Let GoML help you build a reliable, scalable, and secure system for AI in clinical documentation.
