Delphina AI
- AI Startup that is trying to build a DS [[AI Agent]]
- The DS Agent should act as a junior DS
My thoughts
- The product should improve the efficiency of a DS at a company
- The product should replace something boring and repetitive for a DS
- What kind of activities slows down a DS?
- Prototyping
- Framing of the question
- Cleaning, transforming and understanding the available data
- Prompting the DS for data that could be useful (externally)
- Integrate the data of various geography and time
- Able to identify pitfalls of the model created
- Able to use the framing of the question to identify relevant error metrics
- Translate error metrics to user experience/business KPIs
- Productionization
- Able to integrate with model deployment systems
- Able to integrate with datalakes/data systems
- Prototyping
Question for Duncan
- Is Delphina trying to replace the DS team or make them efficient because that is going to be a hard line to tow
- Should they be selling to companies/industries without DS teams but want to make data based decisions
- What kind of problem can Delphina 100% do and replace, help with and not even try to touch?
Process
- Call with cofounder - Duncan
- Technical call with Pedro, given a problem, talk through the tooling/libraries
- Onsite few hours with Delphina demo, interview with Duncan again and ML/whiteboarding session to show thought process