Different models for different jobs
Large summary batches can use lower-cost open-source models. A frontier model then prepares the final Romanian report as structured data that the product can validate and display.
All case studies
Client delivery case study
Building financial reports across 3M+ company records
Topchecks turns a large company dataset into fast verification and risk information. SET Innovation contributed the product architecture, data workflows and an AI module built for cost-aware reporting.
Financial dataAI integrationSaaS
Case study reviewed September 2026
The requirement
Businesses needed to check company information and understand credit and risk signals without manually assembling records from a large financial dataset.
Reports had to remain fast and readable, including when users requested AI-assisted summaries and company-history analysis.
Contribution made explicit
Our role and direct contribution
Aurel “Simi” Simion worked on the architecture and implementation of the platform and its AI processing module.
The contribution covered data-processing services, report generation, product integration and the model-routing decisions used for AI-assisted output. This was client delivery work. The operating figures below describe the Topchecks platform, not a company-owned SET Innovation product.
Constraints and design decisions
The AI feature had to fit the economics and reliability needs of an existing financial product, not behave like a standalone chat demo.
History adds context
Company-history analysis helps the report explain reliability over time instead of reducing a business to one current number.
Reporting stays a product feature
The AI module returns structured output for the application. Users receive a consistent report rather than unreviewed text pasted from a model.
What the work delivered
- Financial verification and reporting across more than 3 million company records.
- Fast report generation supported by services matched to each data workload.
- A separate AI processing module that can route high-volume and final-report work to appropriate models.
- Romanian-language structured output designed for the existing product interface.
Results with context
Operating evidence
The figures below are platform-reported operating metrics. They show the environment in which the work runs, not a guarantee of the same scale or speed for every implementation.
3M+
company records
Dataset covered by the financial reporting platform.
308k+
recent monthly checks
Displayed by the platform as verifications during its most recent reported month.
<1 sec
report generation
Documented platform response at its operating scale.
Platform figures were reviewed in September 2026. The 308k+ figure was presented by Topchecks as the latest monthly total; no independent audit or universal performance promise is implied.
Inspect the live Topchecks platformHave a reporting workflow that needs AI?
Start with the source data, the decision people need to make and the cost of a wrong answer. The model choice comes after that.