Product Analyst, Conversational AI

IntegriChain's Product Management team is hiring a Product Analyst for Prism, our conversational analytics platform. Prism lets people who work in market access at life sciences companies ask questions about their data in plain English and get answers in seconds, instead of sending the request to an analyst and waiting. Behind the scenes, Prism runs on semantic layers we build from our raw SQL data. Those layers are what let it take a business question and map it to the right numbers. The Product Analyst makes sure the questions people ask actually line up with the answers Prism gives back. You'll work every day with Product Management, Data Science, and Engineering to keep those answers accurate and worth trusting.

This isn't a development role, and it isn't about building reports or dashboards. It's an analysis job for someone who's technical but isn't a software engineer. You'll use Snowflake Cortex, along with tools like Claude and Codex, to review and improve the prompts business users type into Prism and the answers it sends back. Much of the work is finding places where the way the data is set up leads Prism to misread a question or return the wrong answer. About half the job is digging into the data. The other half is talking with business owners, certifying that queries return the right results, and writing up what needs to change.

What you'll do

• Prompt review: Go through the prompts business users type into Prism and improve them with Snowflake Cortex, Claude, Codex, and similar tools, so their questions get read correctly against the semantic layer.

• Query certification: Check the queries users run and the answers Prism gives back against the source data, then sign off so the business can trust the results.

• Quality: Keep an eye on what users ask and what they get back, and flag repeat errors, gaps, or misreads so they get fixed.

• Semantic layer: Work with Data Science and Engineering to improve the semantic layers built on our raw SQL tables, including taxonomies, metadata, and labeling, so Prism understands the data better.

• Data analysis: Dig into how pricing, rebates, chargebacks, and other data connect, and use what you learn to shape how Prism represents it.

• Working with the business: Talk with business owners about the questions they need answered, and turn those conversations into better prompts, queries, and data structure.

• Requirements: Write clear technical requirements for tuning the model, restructuring data, or improving the product, based on what your reviews turn up.

• Documentation: Write down where the data comes from, the business logic behind it, and how it fits together, so people understand and trust Prism's answers.

• Roadmap input: Speak up for data quality during Prism planning, so priorities reflect what the data actually shows.

• QA and communication: Set up answer-accuracy test cases with QA each sprint, and help put together release notes, briefings, and presentations that explain how Prism works.