Suggestions
Soren Larson
Data products
Soren Larson is a data professional with a rich background in data/product problem-solving across various industries including ad tech, capital markets/data, private equity, and retail. Larson has a passion for leading and mentoring data science teams, especially in fostering growth opportunities for junior colleagues.
He holds a diverse educational background with degrees in SM, Computational Science & Engineering from Harvard University, Mathematics from Columbia University in the City of New York, and BA in Economics, Mathematics, and Music Theory from Swarthmore College.
Larson has been associated with several prestigious organizations in different capacities, including as an Advisor at Primary Venture Partners, a Limited Partner at The Fund, a Data Science professional at OpenStore, a Distinguished Data Scientist at Walmart, Chief Data Scientist at Consumer Edge, Director of Artificial Intelligence at Pypestream, a Founder of a Startup, Data Scientist at Integral Ad Science, Research Associate at Columbia Business School, and Research Assistant at The University of Chicago Booth School of Business.
Highlights
Our inaugural Identity Con event was a hit online this past spring. What could make it better?!
Being in-person in Chinatown in New York.
A lot’s happened since the last Identity Con
- cookies are dead??
- ai costs down >40x
- sticky enterprise AI adoption
More details to come in the following weeks - expecting fascinating discussions and demos!!
See you in NYC?
AI is creating new economics for personalization.
Personalization used to be a scale intensive business: facing fixed costs
- hire a ds team
- instrument activity
- train models
before we’d see any results.
Since the marginal growth of conversion rates and spend in any instance of personalization can be low, it’s best to apply personalization everywhere to recoup the fixed costs.
Gemini 1.5 Flash + Llama 3.1 8b produce recommenders that now cost less on the margin than Algolia or GCP Personalize.
These models are available serverlessly: no fixed costs to set up.
We’ve heard from customers that existing personalization services feel expensive.
GPT4-level model costs have fallen 240x since they launched. A16z’s Martin Casado expects marginal costs to go to 0.
This week we benchmarked costs of personalization / check it out below.
