Scaling Carbon Emissions Data Mapping for Code Gaia: AI-Powered Bulk Import Across All Emission Categories
Extended an AI-powered bulk import workflow from financial carbon data to all emission categories, so customers can upload any carbon emissions dataset at scale.
Overview
Code Gaia's customers already relied on an AI-powered bulk import tool to map large spreadsheets of financial activity data into carbon emission calculations automatically. DataMax partnered with Code Gaia to extend this proven workflow, called Activity Data Mapping, so that carbon emissions activity data across all emission categories can be uploaded, AI classified, and automatically calculated at scale.
Challenge
The existing import flow only recognised financial activities for carbon accounting, forcing customers with multi-category emissions datasets into manual, row by row data entry. This was a major source of friction for mid to large ESG customers managing thousands of carbon emissions data points across reporting periods, and one of the most frequently requested improvements from both prospects and existing clients. Mapping a single large dataset by hand could take days to weeks of manual effort. The challenge was compounded by some emission categories having very large taxonomies of distinct activity types, and by the need to recognise mixed units and locations within a single uploaded file.
Approach
We extended Code Gaia's existing AI classification pipeline so it can first identify each row's emission category, then match it to the correct activity type using a classification method suited to that category's size, ranging from direct matching for compact categories to AI assisted candidate search for categories with very large taxonomies, before resolving units and locations automatically wherever possible.
Solution
DataMax designed and deployed a multi stage AI pipeline, powered by Amazon Bedrock, that extends Code Gaia's bulk import workflow to carbon emissions data across every category. The pipeline automatically identifies each row's emission category and activity type, using lightweight classification for smaller categories and AI assisted candidate retrieval for categories with very large taxonomies. It then resolves the reported unit and location automatically, falling back to safe defaults when a confident match cannot be found, and feeds the completed data straight into Code Gaia's carbon emissions calculations, with no changes to the underlying data model. What used to take customers days to weeks of manual mapping now runs as an asynchronous AI process completed in a matter of hours. The rollout is staged and feature flagged, starting with pilot customers before expanding to the full customer base.
Results
Ready to accelerate your AI journey?
Let's talk about your data and AI challenges. We'll help you build the right strategy and execute with speed.
Get in Touch