Local AI for confidential data analysis, forecasting, process optimisation and software development – with expert guidance at critical decision points.
IAP sets up a locally running AI model on Apple hardware for your task. It uses the agreed analysis and development tools. You define which steps run automatically and which need approval for your workflow.
For data analysis and forecasting, AI uses IAP’s databionic toolbox (dbt). Software development follows practices developed over 25 years of IAP projects. For process optimisation, modern, locally running AI models analyse workflows and assess possible improvements.
Indicative price: EUR 45,000 excluding VAT. The specific scope and final price will be set out in an individual quotation.
Enquire about local AISchematic target workflow with recurring opportunities to intervene
Human-in-the-Loop
You define the decision points in advance. At those points, experts can correct the data selection, change an assumption or redirect the analysis. AI then continues with the approved data and instructions.
IAP contributes Prof. Dr. habil. Michael C. Thrun’s research expertise in explainable AI and interactive knowledge discovery. His publication on his habilitation describes how expert input at critical decision points helps identify relevant structures in data and derive understandable explanations. This research informs the design of the local AI solution.
Thrun, M. C. (2022): Identification of Explainable Structures in Data with a Human-in-the-Loop. KI – Künstliche Intelligenz (German Journal of Artificial Intelligence), Vol. 36, pp. 297–301. Springer. DOI: 10.1007/s13218-022-00782-6 ↗
Reference configuration
Proposed reference configuration: Apple Mac Studio with M5 Ultra, 512 GB unified memory and a 2 TB SSD. The memory accommodates local model weights and analytical data. The specific configuration is determined by your task, memory requirements and economic considerations.
A multilingual, locally running AI model coordinates work, calls approved tools and explains results. Model selection, licence and configuration are determined for each project. Suitability for your task is assessed against agreed test cases.
For data analysis and forecasting, IAP’s databionic toolbox (dbt) provides the analytical methods and calculations. For process optimisation and software development, AI is connected to the tools and interfaces agreed for the task.
A local browser interface shows the connected data, work progress and results. It also provides the controls for intervening in the analysis.
Proposed reference configuration; specifications and suitability are determined for your assignment. Technical specifications at Apple ↗
Applications
For critical business data that must not leave the company, IAP sets up analysis within your own infrastructure. AI uses the databionic toolbox (dbt) to prepare data, investigate patterns and anomalies, and present traceable results. Domain experts assess relevant structures, correct assumptions and guide the analysis.
AI uses dbt to analyse time series, produce forecasts and compare scenarios, for example for demand and available capacity. Your experts contribute knowledge of exceptional events and changing conditions. Forecasts and assumptions are reviewed before informing planning decisions.
AI examines documented workflows, prepares bottleneck analyses and improvement options, and compares them against agreed objectives and constraints. Process owners review proposals at defined decision points.
Precisely specified Jira tickets or Kanban tasks are translated into implementations in R, Python or C++. AI performs implementation, testing and documentation according to project-specific requirements. Experts guide architectural decisions, review changes and approve results against predefined acceptance criteria.
Software development to defined standards
The approach draws on methods and quality criteria from 25 years of IAP project practice. It also builds on the expertise of Prof. Dr. Alfred Ultsch ↗ and Prof. Dr. Michael Thrun ↗ in translating scientific methods and research findings into reproducible, quality-assured software ↗.
For AI-assisted implementation, this experience is translated into concrete project-specific requirements: clear specifications and acceptance criteria, consistent code structures and interfaces, versioned changes, verifiable tests and traceable technical documentation.
Each change is linked to its requirement. Code, tests and documentation are versioned together so reviewers can check what was implemented and how it was tested.
The intended scope includes hardware and technical setup, model installation and configuration, integration of the agreed tools, testing against agreed test cases, documentation and onboarding.
Defining and configuring project-specific expert intervention and approval points is the customer’s responsibility and is not included in the indicative price. IAP can provide this service under a separate agreement for an additional fee.
Following acceptance, the solution is handed over to your IT team or operating partner. Ongoing operation and continuing support by IAP are not part of the service.
Describe your workflow and local processing requirements. We discuss the scope needed as the basis for an individual quotation.
