89 集
The Data Scientist Show Daliana Liu
-
- 科技
-
-
4.7 • 3 則評分
-
A deep dive into data scientists' day-to-day work, tools and models they use, how they tackle problems, and their career journeys. This podcast helps you grow a successful career in data science. Listening to an episode is like having lunch with an experienced mentor. Guests are data science practitioners from various industries, AI researchers, economists, and CTOs of AI companies. Host: Daliana Liu, an ex-Amazon senior data scientist with 180k followers on Linkedin.
Join 20k subscribers at www.dalianaliu.com to learn more about data science, career, and this show. Twitter @DalianaLiu.
-
Why data scientists are tired, six real data scientists' frustrations - The Data Scientist Show #089
Daliana interviewed 6 data scientists from her meetup in New York City. It's a unique episode where you get to hear the real frustrations of data scientists. We talked about struggles working in healthcare, finance, data quality and AI, how to advocate for yourself, and align with your managers.
Subscribe to Daliana's newsletter on www.dalianaliu.com for more on data science and career.
Daliana's Twitter: https://twitter.com/DalianaLiu
Daliana’s LinkedIn: https://www.linkedin.com/in/dalianaliu/ -
Why 80% of A/B tests fail, how to 10X your experimentation velocity - Kristi Angel - The Data Scientist Show #088
Most experimentations fail, Kristi Angel shares her expertise on scaling experimentation and avoiding common A/B testing pitfalls. Learn five things that can help boost test velocity, designing impactful experiments, and leveraging knowledge repos. (Chapters below)
Kristi Angel’s LinkedIn: https://www.linkedin.com/in/kristiangel/
Subscribe to Daliana's newsletter on www.dalianaliu.com for more on data science and career.
Daliana's Twitter: https://twitter.com/DalianaLiu
Daliana’s LinkedIn: https://www.linkedin.com/in/dalianaliu/
(00:00:00) Intro
(00:01:26) Why do most experimentations fail?
(00:07:05) Mistakes in choosing metrics
(00:10:05) Is revenue a good metric?
(00:13:18) Split metrics in three ways
(00:15:10) Daliana's story with too many category breakdowns
(00:16:59) What makes the best data science team?
(00:19:24) Data scientist work in silo vs in a data science team
(00:21:15) Building a knowledge center
(00:23:40) Example of knowledge center; nuance of experimentations
(00:26:09) How many metrics and variants?
(00:30:56) How to reduce noise - CUPED
(00:33:01) Future of A/B testing
(00:38:33) Q&A: Low statistical power -
From physics PhD to data science leader, unexpected challenges in survey data, Python vs R, EDA best practices, building MLOps toolkit - Julia Silge - The Data Scientist Show #087
Julia Silge is an engineering manager at Posit PBC, formerly know as R-studio, where she leads a team of developers building open source software MLOps. Before Posit, she finished a PhD in astrophysics, worked for several years in the nonprofit space, and was a data scientist at Stack Overflow where some of her most public work involved the annual developer survey. We talked about MLOps tools, challenges in survey data, text analysis, and balancing her interests in data science and engineering.
Subscribe to Daliana's newsletter on www.dalianaliu.com for more on data science and career.
Daliana's Twitter: https://twitter.com/DalianaLiu
Daliana’s LinkedIn: https://www.linkedin.com/in/dalianaliu/
(00:00:00) Introduction
(00:00:56) Getting into data science
(00:04:50) Transition from data centers to engineering manager
(00:14:04) Common challenges in tool development
(00:17:38) Challenges with survey data
(00:26:47) Engineering skills for data scientists
(00:28:59) Balancing roles
(00:34:49) Developing skills in Exploratory Data Analysis (EDA)
(00:39:19) Python vs. R for data analysis
(00:44:40) Exciting aspects in career and personal life -
Why he created Pandas, the future of data systems, why he left his CTO role to become a chief architect - Wes McKinney - The Data Scientist Show #086
Wes McKinney is the co-creator of pandas library and he is the cofounder of Voltron data. Currently he is a principal Architect at Posit and an investor in data systems.
Daliana's Twitter: https://twitter.com/DalianaLiu
Daliana’s LinkedIn: https://www.linkedin.com/in/dalianaliu/
Wes' LinkedIn: https://www.linkedin.com/in/wesmckinn/
(00:00:00) Introduction
(00:00:44) How Pandas Started
(00:06:40) Voltron Data
(00:10:03) Benefits of Easy-to-Use Data Tools
(00:13:20) The Rise of New Data Tools
(00:18:07) Choosing Tools: Vertical or Flexible?
(00:23:01) Big Models and Data Tools
(00:29:29) Challenges in Building a Product
(00:31:28) Becoming a Top Architect
(00:34:55) Missed Aspects of Previous Roles
(00:39:04) A Busy Week: Advising, Designing, Investing
(00:43:42) Improving Open Source
(00:45:24) How to Decide What to Work On
(00:46:28) What he’s learning now
(00:47:56) Excitement in Career and Life
(00:48:29) Using ChatGPT for Learning
(00:50:27) Future Impact Goals -
From financial analyst to director of analytics, how to get promoted quickly, 7 elements of influence - Christopher Fricker - The Data Scientist Show #085
Christopher Fricker is a senior director in analytics and BI at Renaissance Learning. He started his career in finance and later became a data science consultant working with Meta, Netflix, and pre-IPO tech companies doing analytics. We talked about the mental models that helped him grow from a finance analyst to an analytics leader.
Subscribe to Daliana's newsletter on www.dalianaliu.com for more on data science and career.
Chris’ LinkedIn: https://www.linkedin.com/in/christopherfricker/
Daliana's Twitter: https://twitter.com/DalianaLiu
Daliana’s LinkedIn: https://www.linkedin.com/in/dalianaliu/
(00:00:00) Introduction
(00:01:46) How to get promoted quickly
(00:08:40) Power vs authority
(00:11:21) First principal thinking
(00:32:34) ROI of a data team
(00:40:29) How to be persuasive
(00:54:52) All Data is wrong
(00:56:22) How he audits the data
(01:00:52) How to make someone help you at work -
Adapters: the game changer for fine-tuning - Geoffrey Angus - The Data Scientist Show #084
I interviewed Geoffery Angus, ML team lead @Predibase to talk about why adapter-based training is a game changer. We started with an overview of fine-tuning and then discussed five reasons why adapters are the future of LLMs. Later we also shared a demo and answered questions from the live audience. Try fine-tuning for free: https://pbase.ai/GetStarted
Geoffrey’s LinkedIn:https://www.linkedin.com/in/geoffreyangus
Daliana's Twitter: https://twitter.com/DalianaLiu
Daliana’s LinkedIn: https://www.linkedin.com/in/dalianaliu/
Daliana's Twitter: https://twitter.com/DalianaLiu
Daliana’s LinkedIn: https://www.linkedin.com/in/dalianaliu/
Geoffrey’s LinkedIn: https://www.linkedin.com/in/geoffreyangus
Try finetuning for free: https://pbase.ai/GetStarted
(00:00:00) Intro
(00:01:19) What is Fine-tuning?
(00:08:18) Utilizing Adapters for Finetuning Enhancement
(00:09:50) 5 reasons why adapters are the future of LLMs
(00:26:34) Common Mistakes in Adapters Usage
(00:28:34) Training Your Own Adapter
(00:32:23) Behind the Scenes of the Adapter Training Process
(00:37:51) Config File Guidance for Fine-Tuning
(00:39:41) Debugging Strategies for Suboptimal Fine-Tuning Results
(00:42:23) User Queries: Creating a LoRa Adapter and Future Support
(00:51:06) Key Takeaways and Recap
客戶評論
熱門科技 Podcast
你可能也會喜歡