From hours of manual neuroscience reports to fully automated insights in minutes

The client's work generates highly granular data: emotional responses measured frame by frame, across multiple KPIs and audience segments.

But behind these insights, the process was entirely manual. Data lived across dozens of Excel files, with no consistent structure.

Each report required analyzing thousands of data points, building charts, interpreting results, and manually creating PowerPoint presentations.

As demand grew, this became a bottleneck. What once worked at a smaller scale was no longer sustainable.

In order to structure, organize, and consolidate fragmented data, Astrafy built a Python-based processing layer to ingest, validate, and standardize all inputs. This included:

• Parsing multiple Excel files

• KPI calculations at scale (confusion, recall, rejection, etc.)

• Applying segmentation logic programmatically across multiple dimensions

On top of the structured data, Astrafy automated the full reporting workflow, enabling end-to-end automation of report generation.

Using Gemini:

• Generate insights and conclusions from KPI data

• Interpret emotional patterns across time

Using Python:

• Create charts and visual assets

• Apply business logic (e.g. KPI thresholds)

• Assemble complete PowerPoint reports

To enable business users, Astrafy built a web interface where users upload a .zip file and receive a ready-to-use report. This turned a technical pipeline into a usable product. Architecture:

• Vertex AI → Gemini integration

• Cloud Run → backend + frontend

• Cloud Storage → file handling

• Report generation reduced from hours to ~2 minutes

• Increased reporting capacity without additional resources

• Fully automated workflow (data → insights → presentation)

• Autonomy for non-technical teams