Today's Guest: Maksim Butsenko

Background: 6 years in Data Science at Bolt

Projects: Real-time pricing, user incentives, growth portfolio optimization

Current Focus: Customer service automation with LLM

In this episode:

* Maksim's journey and experience at Bolt

* The structure of the ML teams

* Cooperation between ML teams and Product teams

* Differences between Research based and Traditional PM

* Challenges and best practices in Research based PM

* ChatGPT for D&D, best ML model, and thoughts on AI automation

Timestamps:

0:00 Intro

1:13 Welcoming the guest

3:16 Scaling Data Science Team at Bolt

5:29 Structure of Data Science Team at Bolt

9:40 Role of Product Manager

11:24 Where does Bolt use ML?

12:54 Does Bolt build in-house or use external tools for ML?

16:10 Progression of Data Science in Bolt

22:13 AI startups

24:28 Collaboration between Data Scientists and Product Managers

29:20 PM guiding the Team

33:33 3 data teams in Customer Support at Bolt

35:44 How Bolt started doing ML

36:06 LLMs at Bolt

37:01 Building Alfred

38:28 How to lead Research teams

43:51 How to plan DS / ML projects?

46:19 Estimating the Impact

50:10 How Bolt Estimated Impact of GPT-4

52:22 One metric at a time

53:44 LLM costs

55:48 Should your company use ML?

59:16 Complexity of doing Research

1:04:20 Research as a "secret sauce"

1:06:50 Expectations of Product Managers

1:09:08 Technical Expectations of Product Managers

1:10:53 Advice to PM in Research team

1:13:08 How Maksim devotes time for Research

1:16:16 Choosing the Right LLM / ML model

1:19:48 Maksim's Favorite ML model

1:20:56 D&D with chatGPT

1:27:14 Will AI automate everything?

Connect with us:

* Maksim Butsenko at LinkedIn: https://www.linkedin.com/in/maksim-butsenko-1b13194a/

* Nikolay Roll at LinkedIn: https://www.linkedin.com/in/nikolay-roll/

This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit www.tpg.ee