Submit offer AI Engineering Advisory — Knowledge Retrieval & Clinical Compliance

Template for the Task Description
  
A good Task Description must include the following:  
 
1. GENERAL INFORMATION: 
Titel of tender: AI Engineering Advisory — Knowledge Retrieval & Clinical Compliance
Your company’s name:  Themison ApS
Address:  Fruebjergvej 3, 2100 Copenhagen Ø, Denmark
CVR nr.:  44224062
Date:  31/03/2026
Deadline for tender:  09/04/2026
For further questions, contact info:  Michael Pavlou, CTO
Upload of tender info: Use link at task description
Phone and e-mail:  +45 91858308 — michael.pavlou@themison.com
 
2. PRESENTATION OF COMPANY: 

Themison is a digital health startup developing an AI-powered platform to support clinical trial operations at site level. The platform focuses on improving patient engagement, operational efficiency, and protocol adherence.

The system is designed to assist clinical trial staff by enabling efficient retrieval, interpretation, and operationalization of protocol-related information. It integrates multiple data sources, including clinical trial protocols, amendments, operational documents, and structured data, to provide context-aware insights and decision support.

The platform currently leverages large language models and retrieval-based architectures (RAG) to process and interpret clinical documents. The goal is to evolve this into a scalable, compliant, and production-ready system that can operate within regulated clinical environments.

Themison is currently entering a co-development phase with clinical sites, where the system will be tested and refined in real-world conditions.

Website: www.themison.com

 
3. DESCRIPTION OF THE TASK UNDER MARKET EVALUATION: 

Themison is seeking an external consultant with expertise in AI architecture and regulatory-compliant system design to support the refinement of its platform.

The objective of this task is to strengthen the system’s architecture, ensure alignment with regulatory requirements, and establish a robust foundation for scaling the product.

The consultant is expected to provide guidance in the following areas:

Retrieval Architecture & Multi-Source Intelligence

Transfer knowledge on hybrid retrieval design (dense + sparse search with metadata filtering) for simultaneously querying across multiple document types and operational data sources
Advise on knowledge graph architecture that encodes relationships between clinical trial documents — how an amendment overrides a protocol section, how procedures map to roles, how changes propagate across dependent tasks

Context Assembly & Report Generation

Advise on designing a context assembly layer that combines retrieved information from multiple sources into coherent, structured prompts enabling situation-aware responses rather than isolated document extractions
Provide guidance on architecting a report generation pipeline: from user request through multi-source retrieval to a structured, source-traceable clinical output document

Compliance, Anonymisation & Safe Deployment

Advise on data anonymisation approaches appropriate for operational clinical trial data processed by an AI system, ensuring no patient health data is exposed or retained
Provide guidance on how to design the system to meet ICH E6 GCP, GDPR, and EU CTR 536/2014 requirements — including audit trail design, data minimisation, role-scoped access, and zero-retention options for external model inference
Advise on an evaluation and validation framework (such as RAGAS or equivalent) that meets the evidentiary standards expected in a regulated clinical environment
All recommendations will be implemented by Themison's internal engineering team. The advisor's role is strictly to provide specialist expertise, architectural guidance, and knowledge transfer that Themison does not currently have in-house.

4. TASK OBJECTIVES AND SUCCESS CRITERIA: 

Objective: 

Acquire the specialist knowledge and architectural guidance needed to advance Themison's AI system to multi-source contextual intelligence, structured report generation, and compliant deployment in regulated clinical environments.

Those capabilities are central to the platform's clinical value.

Success is measured by:

A recommended hybrid retrieval architecture defined and documented, covering multi-source querying across at least 3 source types simultaneously
A knowledge graph schema recommendation documented, covering core clinical trial entities, relationships, and amendment propagation logic
A recommended context assembly and report generation pipeline architecture documented, covering the full flow from user request to structured clinical output with source citations
Documented guidance on anonymisation approach, compliance architecture, and audit trail design meeting GDPR, ICH E6 GCP, and EU CTR requirements
Citation accuracy on a defined internal test set improves measurably following implementation of the advisor's recommendations, with a target of ≥90% correct source attribution

 
5. BUDGET OG SPECIFICATION OF AN OFFER: 
We expect a written offer to include at least: 
• Date of submission of offer 
• A brief presentation of the bidder, stating the CVR number and contact details. If relevant, with references and history 
• Bidder's proposal for solving the task 
• Specification of the price for solving the task: Estimated number of hours and hourly rate, or a fixed price, aligned with the expected scope of approximately 70-90 hours of senior advisory work and a maximum budget of DKK 47,000 ex VAT or €6,300 ex VAT
• Discount, if relevant 
• Timeframe and end date 
• Conditions for the offer, if any 
 
6. BACKGROUND FOR THE TENDER: 
Beyond Beta is subject to a number of requirements for good, healthy financial management, including documentation that the agreed price for external purchases is an expression of the market price. This tender is part of these requirements. 
We emphasize that the bidder must only make an offer on the requested task.  
Services of executing or implementing nature cannot be approved  
The winning bid is chosen based on an assessment of the best correlation between price and quality