Projects

Chat and agent platform

Venice is a privacy-first generative AI platform offering text, image, and video models with no data retention.

I own the roadmap for the Agentic Systems squad and work closely with engineering, design, and leadership.

• Agent platform. I drive the roadmap for agentic chat and Minds, Venice’s platform for publishing AI agents, from tool design and approval flows to creator monetization.
• Model strategy. I manage the text, image, and video models behind our agents. I evaluate new models, move users to new defaults, and manage model retirements so creators’ agents keep working without surprises.
• Privacy-first integrations. I scope how agents connect to third-party apps for email, documents, and storage while keeping user credentials off our servers, and I make those tradeoffs clear to users.
• Agent safety. I assess the security risks of agents that can act on a user’s behalf, including prompt injection, sandboxing, credential theft, and approval design, and turn them into product requirements.
• Strategy. I write strategy docs on where the product goes next, such as whether and how Venice should build a desktop agent.

Ads reporting and measurement

I owned Spotify’s ads reporting and measurement, from the underlying data pipelines to the tools advertisers used to understand campaign performance.

Ads reporting

My teams owned the full reporting pipeline, from ingesting and cleaning event data to delivering campaign reports to advertisers.

What we built for measurement

• Identity exchange. We built an ML-powered identity exchange that used billions of signals to resolve identity for ads measurement. It gave us a way to connect events at the user level so they could be used for attribution.
• User-based attribution. We built tools that connected ad exposure with later actions at the user level. That gave advertisers a clearer view of which outcomes followed exposure to their Spotify campaigns.
• Incrementality. We built products that helped advertisers estimate the additional outcomes their campaigns generated beyond what would have happened without the ads. That helped them judge the effect of their spend.
• Brand lift. We built ML-powered brand lift tools to help advertisers understand how campaigns influenced what people knew or thought about their brand.
• Clean room integrations. We built integrations that let advertisers use their own data alongside Spotify’s in a controlled environment to answer measurement questions.
• Mobile measurement. We built tools to connect Spotify ad exposure with outcomes in mobile apps. Advertisers could see which app activity followed their campaigns.
• Privacy-first measurement. We developed ways for advertisers to measure campaign performance while limiting how much user data needed to be shared in the process.

Conversion measurement

Connecting what advertisers measure in Google Analytics with how they bid and optimize in Google Ads.

The product problem

A conversion signal has to support two uses. It needs to help an advertiser understand performance, and it needs to be useful for campaign optimization.

My role

I led product strategy for Analytics and conversion measurement, with a focus on linking measurement capabilities to Google Ads bidding and optimization.

What we shipped

We shipped attribution and experimentation features for measuring advertising performance, along with causal inference and privacy-safe signal systems that supported conversion measurement and ads optimization.