Product Manager on an AI risk analytics platform, sitting between data science, engineering and the partners who have to trust the output.
Risk work at a Big Four firm runs on judgment that lives in experienced people's heads. The platform I manage puts analytics and AI underneath that judgment: surfacing anomalies, scoring exposure, and turning weeks of manual review into hours of directed attention.
The hard part is not the model. It is that the output goes to people whose signature carries professional liability. A partner does not accept "the model says so". Every number needs a trail back to evidence, every screen has to survive the question "how do you know that", and adoption only happens when the tool is more defensible than the spreadsheet it replaces.
Data scientists speak in precision and recall. Partners speak in exposure and defensibility. The product exists in the space between those two vocabularies, and so do I.
The users are experts with twenty years of habit. Shipping the feature is a third of the job. Getting a risk team to trust it with a live engagement is the other two thirds.
In a regulated environment, an answer without provenance is a liability. I spend real roadmap on lineage, audit trails and explainability, the unglamorous parts that decide whether the platform is allowed in the room.
I like comparing notes on evidence, explainability and getting experts to change their workflow.
venkatesh@venkateshgardas.com