SALI UN

International sustainability reporting runs on an indicator that member states have to report against, and the underlying evidence arrives as thousands of corporate reports in no consistent shape. Aggregating that into something a body can publish is not a data problem at the end, it is a data problem at every step from ingestion onward.

The ingestion, categorisation, analysis, aggregation, and reporting pipeline, with raw data in object storage and metrics and aggregates in a document database.

A framework knowledge base spanning the sustainability disclosure and reporting standards the initiative works to, and roughly two hundred and thirty environmental, social, and governance data points organised into reusable categories so scoring, benchmarking, heatmaps, and compliance tooling all draw on one backbone rather than each defining its own.

Predefined and custom reporting, an interactive dashboard API, and role and module based access control across the whole surface.

A pipeline of specialised agent tiers, each with one job, so a change to how documents are filtered cannot disturb how metrics are validated.

A knowledge base holds the frameworks themselves and serves them through vectorised retrieval. A categorisation tier filters incoming documents for relevance before anything expensive touches them, which is what keeps the cost of the pipeline proportional to the useful documents rather than to all of them. An analysis tier extracts and validates metrics. An aggregation tier rolls them up by company, sector, and region. A bridge tier accepts input from the wider product and evaluates it against the same standards.

Each module is self contained, with its own routes, services, schemas, and helpers, so modules are developed, tested, and scaled independently.

Storage moved from PostgreSQL to a document database as the work went on. The records being kept are extracted metrics whose shape is set by whichever reporting framework produced them, and a schema that has to be migrated every time a framework gains a field is a schema fighting the problem.

Engineer. The agent pipeline, the knowledge base and retrieval, the aggregation layer, and the reporting and dashboard surfaces.

Contributed to sustainability technology supporting international organisations, including UNCTAD and UNEP.