Juergen Dietrich

- TRUST: Design and development of the TRUST multi-agent LLM pipeline for automated democratic discourse analysis (democracy-intelligence.de):

- Heterogeneous ensemble: relevance filter, web-search-grounded fact-checker, three epistemic advocates & supervisor, iterative consensus logic – Bilingual (DE/EN), multi-provider LLM support (OpenAI, Anthropic, Google); scoring across Logos, Ethos, Pathos using a Nutri Score-inspired A–E label scheme

- Three publication preprints published (April 2026), one is in preparation (May 2026)

Additional consulting & development activities:

- Development and provision of medDetect — AI system for automated detection of pharmaceutical terms (drug, indication, adverse event, dose, route) in free text; commercially deployed.

- Contribution to CRO qualification audits for AI-based service providers in pharmacovigilance and data management.

- Consulting: AI project management, validation of AI systems in controlled environments, training data development for client-specific AI systems.

- AEDE: Design, development, and validation (as Subject Matter Expert) of AEDE (Adverse Event Detection Engine) Bayer's production ML system for automated detection of adverse events, indications, and drugs from unstructured sources (Twitter/X, PubMed, customer calls)

- Development of a dataset and NER models for pharmacovigilance and Medical Affairs;

- novel fractional k-fold cross-validation methodology for training data sufficiency in GxP environments

- Automatic assignment of WHO Drug Dictionary C3 formats to Bayer's drug master data

- Development of new methodologies for AI deployment in regulated GxP environments (2 patent applications, 2021).

- Author of 3 peer-reviewed Drug Safety publicatons on LLM performance/reproducibility and NER corpus design (2022– 2024).

- Bayer's representative in IMI WEB-RADR (EU initiative for social media use in pharmacovigilance); Four Drug Safety publications.

- Initiated and managed innovation projects:

-- Signal detection for social media platform,

-- NLP for benefit analysis,

-- Deep learning for case processing and signal detection (PoC).

- Budget management, business case creation, leadership communication, stakeholder management across global PV.

- System implementations:

-- Empirica Signal Detection,

-- CAPA/Root Cause Analysis,

-- xEVMPD/eRMR mapping,

-- EVWEB E2B autotriage,

-- E2B R3 support for Russian health authorities.