Case Study

Building AI-Powered Compliance Tools for Life Sciences

Case Study

Building AI-Powered Compliance Tools for Life Sciences

Client

Canopy Life Sciences

73 Strings

https://www.hellocanopy.com/

www.73 Strings.com

Industry

Healthcare & Life Sciences

Finance/Fintech
Alternative Asset Valuations
Enterprise SaaS

Services

Analysis & Design
Source Document Vectorization
Prompt Engineering
Retrieval Augmented Generation (RAG)

Discovery Research
UX/UI Design
Design System
Front-End Engineering

Project Duration

6 Months

6 months

Problem

Canopy Life Sciences, a provider of advanced services for regulatory and promotional content review for the pharmaceutical industry, sought to enhance the speed and consistency of their compliance workflows. Their clients in life sciences face increasing demands for efficient, compliant claims substantiation and document validation processes - often under tight timelines and strict FDA regulations. Manual review processes were time-consuming, prone to inconsistencies, and hard to scale. Canopy’s goal was to introduce an AI-driven system to reduce the burden on reviewers, improve metadata accuracy, and ensure regulatory adherence — all while maintaining strict data segregation and customer-specific models.

Our Vision

Roko Labs partnered with Canopy to co-design and deliver a scalable, compliant AI system that works seamlessly within the Veeva Vault environment. Working as a fully integrated product and engineering team, we leveraged Canopy’s existing APIs and regulatory expertise while applying our own experience in enterprise AI infrastructure. The team followed a sprint-based, iterative delivery process with deep collaboration on architecture and model design. Key decisions were reviewed jointly to ensure compliance, performance, and customer-specific isolation across language models and tenants.

Problem

Our Vision

Problem

Our Vision

Project timeline showing discovery, system design, and implementation phases across three sequential stages.

Solution

Roko labs developed a custom AI-powered compliance orchestration system to automate the Medical, Legal, and Regulatory (MLR) review. The system enables business users to submit documents, which are then analyzed against country-specific regulation guidelines (e.g., FDA 202.1) and product reference documents. Annotated results are returned, highlighting compliance issues for expert review.

Solution

Software dashboard displaying document processing status and compliance test results with passed, failed, and in‑progress metrics.

People working at desks in a bright office environment, using laptops and notebooks during a collaborative work session.

Digital compliance workflow visualization showing document files processed through a central system with a passed compliance status.

Execution Details

Roko’s AI engineers designed a robust system architecture centered around a fine-tuned single model, incorporating Retrieval Augmented Generation (RAG). This approach enabled the system to effectively process domain-specific knowledge, ensuring precise extraction of key information from complex medical and regulatory documents.

Execution Details

MLRAI Admin UI  Application for administrators to configure regulatory guidelines, upload product references, and view results.

MLRAI Web API  Central hub for all requests, secured with Basic Auth. Connects Document and Admin UI to backend processing, queues jobs, and persists results.

MLRAI Process Service  Service that orchestrates the document analysis pipeline: content extraction, AI analysis, annotation, and persistence.

Pagination & OCR Lambdas  Split PDFs into pages (mlr-ai-paginate) and convert page images into structured text and spatial data.

System architecture diagram showing document upload, AI compliance platform analysis, and LLM integration for automated document review.

Technology

The system is powered by a suite of specialized AI components that work together.

Technology

The system is powered by a suite of specialized AI components that work together.

Distributed AI Workflow  We delivered a system where OCR, retrieval, and language models each play a specialized role, working in coordination to provide layered analysis without relying on a single monolithic service.

Config-Driven Intelligence  All AI behavior is governed by prompts, guidelines, and references that can be updated instantly—no retraining or redeployment required—significantly reducing maintenance overhead.

Orchestration & Control Layer  A centralized process service manages sequencing across SQS queues, Lambdas, and APIs, ensuring reliable execution, fault tolerance, and consistent results.

Agile Testing & Iteration  Through Swagger APIs and admin tools, prompts and workflows can be quickly adjusted, tested, and validated, enabling fast cycles of improvement.

Cloud-Native Scalability  The platform leverages AWS-managed services (ECS Fargate, Lambda, S3, Bedrock, OpenSearch, PostgreSQL) to scale automatically and adapt easily to new regulations, products, and AI models.

Workflow diagram illustrating an end‑to‑end document processing system, from user submission through AI execution, persistence, and final review.

Testimonials

“I'm very impressed with the team. They have picked up on the problem domain very quickly, and I find them incredibly easy to work with. Glad we engaged with you. Shout-out to the AI team -- they gave us a very professional and detailed demonstration of the workings of the AI product. It was exactly what I had hoped for!”

Kevin Collins,
Chief Technology Officer

Canopy Life Sciences

“Our internal team had great feedback and was very positive about the improvements. I’m really pleased with how it’s all turned out!”

Bouchra Lahnin,
Project Manager,

Canopy Life Sciences

Need a custom AI solution?
See how we turn ideas into real products.

Learn more

More Case Studies

CASE STUDY