Expert AI Implementations
AI Strategy and Roadmap
- Evaluation of client’s PV environment with respect to intelligent automation, knowledge processes, data management and technical capacities.
- Tailored strategy and roadmap for more automated and AI-driven/-assisted PV processes including data acquisition (monitoring, intake), case processing or signal detection.
- Transition by initialising and driving relevant Proofs of Concepts, implementation projects and corporate change initiatives.
- Typical improvements might include:
- Monitoring of relevant data sources - literature, social media, mail inboxes etc.
- Exploratory Data Analysis, data/text preparation and pre-processing
- Entity recognition and pre-coding (MedDRA PTs, products), AERs and duplicate detection/classification, seriousness suggestion, case summarization
- Replacing classic predictive ML processes for case classification with on-prem Generative AI backend
- Simulations, visualizations, modeling temporal relations and causality assessment
- Better explainability and validation compliance into AI solutions
- Better tuned models (hyper-parameters) for regression, classification, deep learning (NLP)
- Productionalizing PV IT operations towards scalable architectures, faster turnaround (CI/CD), continuous improvements and Machine Learning PV IT operations (MLOps covering train/test/validation/production pipelines, etc.)
Generative AI Implementations
- Onboarding the GPT Prompt interface for client's Safety Department office workers
- Evaluating the best strategy for front-end office GPT use considering various options e.g. using mainstream Prompt vendors, 3rd party solutions and prompts plugins, subscribing to vendor AI platforms (e.g. MS365 Co-pilot, Azure OpenAI, automation agents etc.), partnering with a specific LLM vendor
- Prompt engineering and fine-tuning
- Designing governance around GPT for safe and compliant Prompt - Gen AI SOP, Prompt Engineer + AI QA roles...
- Incorporating Generative AI into products and IT architectures
- Prompt as a Platform and relevant Prompt configuration management
- Using Vector databases to manage critical company knowledge
- Implementing Retrieval Augmented Generation pattern for compliant knowledge management system that understands company documents, has knowledge of own QMS, case and risk management practices, and is able to support PV workers throughout multiple PV processes (PV AI Co-pilot).
