AI Portfolio Podcast

Ramsri, an AI expert with 12+ years of experience in AI and product development, specializing in computer vision and NLP. Ramsri is the founder of Questgen.ai and Supermeme.ai, innovative NLP SaaS apps. He also teaches NLP and AI SaaS courses at LearnNLP Academy and shares insights on Medium, Twitter, and YouTube.

📲 Ramsri Goutham Golla Socials:
LinkedIn: https://www.linkedin.com/in/ramsrig/
Twitter: https://twitter.com/ramsri_goutham

📲 Mark Moyou, PhD Socials:
LinkedIn: https://www.linkedin.com/in/markmoyou/
Twitter: https://twitter.com/MarkMoyou

📗 Chapters
00:00 Intro
02:43 Bootstrapping Two Companies
05:09 PhD and Entrepreneurship
08:11 Overview of Companies
09:53 Memes
12:23 Bootstrapping Startups
22:27 Minimalistic Philosophy for Startups
25:47 Building Distribution through Personal Brand
34:39 Saas vs Micro Saas
40:15 Happy Place
47:51 Major Components of AI saas apps
54:04 Open Source Models vs API
58:26 Getting Featured in Google
01:08:17 Coming back to India
01:17:36 Takeaways from Silicon Valley
01:32:48 Company of one
01:35:39 Number One on Product Hunt
01:45:25 Generalize High Cognition Barrier
01:53:15 Mistakes in building Saas
01:59:17 Unfair Advantage
02:01:53 Advice for launching micro saas apps
02:05:09 If you have to start over
02:08:36 Advice for high schooler, college student and professionals
02:22:17 Rapid Round

Serg Masís, a renowned Data Scientist in the agriculture sector, with a multifaceted background in entrepreneurship and web/app development. Serg is also the author of the acclaimed book "Interpretable Machine Learning with Python."

📲 Serg Masís Socials:
LinkedIn: https://www.linkedin.com/in/smasis/

📲 Mark Moyou, PhD Socials:
LinkedIn: https://www.linkedin.com/in/markmoyou/
Twitter: https://twitter.com/MarkMoyou

📗 Chapters
00:00 Intro
03:02 Interpretable Machine Learning System
11:06 post-hoc interpretability
12:57 Quality of Interpretability bounded by the breadth of inputs
18:52 EDA is more than just exploring the data
21:12 Interpretable Models
27:19 Upper Limit
28:59 Causal reasoning
35:02 Domain Knowledge in Causality
37:42 Causality on existing and generative data
40:42 AGI and Simulation
47:33 LLMs to Interpret Data
52:25 Interaction with Machine Learning Systems
54:53 Modeling in EDA
57:34 Training Data Mixtures
01:02:26 Impact of Generative AI
01:06:05 Mixture of Experts
01:12:46 Serg's upcoming book
01:18:54 Building Product Mindset
01:23:59 Barrier of Entry for Data Scientists
01:28:01 Advice for Book Writers
01:33:43 Career Optimization Function
01:37:45 Web Development Background
01:42:24 Advice for High Schooler, College Student and Professionals
01:47:57 Book Recommendations
01:49:26 Rapid Round

Sanyam Bhutani, a leading figure in the data science community. Sanyam is a Sr. Data Scientist at H2O.ai, with previous tenures at Weights & Biases and H2O.ai, and an International Fellow at fast.ai. As a Kaggle Grandmaster, his contributions to the field are well-recognized and highly respected.

Sanyam delves into the nuances of fine-tuning and optimizing Large Language Models (LLMs). He provides a detailed exploration of the current state and future potential of LLMs, breaking down their architecture and functionality in a way that's accessible to both newcomers and seasoned data scientists. Sanyam discusses the importance of fine-tuning in enhancing the performance and applicability of LLMs, providing practical insights and strategies for effective implementation.

📲 Radek Osmulski Socials:
LinkedIn: https://www.linkedin.com/in/sanyambhutani/
Twitter: https://x.com/bhutanisanyam1?lang=en

📲 Mark Moyou, PhD Socials:
LinkedIn: https://www.linkedin.com/in/markmoyou/
Twitter: https://twitter.com/MarkMoyou

📗 Chapters
00:00 Intro
02:46 200 days of LLMs
06:16 Venture Capital
08:40 Setting Goals in Public
09:45 Fine tuning Experiment
14:02 Kaggle Grandmasters Team
15:55 Doing Challenges & Reading Research Papers
17:47 Hardest topic to learn in AI
19:05 Are you afraid to ask stupid questions?
20:43 Learning how LLMs work
22:54 Academic vs Product First Mindset
27:51 Training or Inference on LLMs
29:15 Favorite LLM Agent
32:10 How to go about learning LLMs?
36:55 Open Source LLMs on Research Papers
37:41 Capability of Modern GPUs
45:48 Journey to H20.ai 
50:07 Why Sanyam stopped podcasting?
56:25 Podcasting Experience
58:39 Top Data Scientists
01:00:19 Advice for New Podcasts
01:03:32 Breaking into Data Science
01:12:23 Career Optimization Function
01:14:02 Making Progress Everyday
01:15:05 Advice for New Professionals
01:17:00 Book Recommendations
01:18:04 Rapid Round

Radek Osmulski, a Senior Data Scientist specializing in Recommender Systems at NVIDIA, brings a wealth of experience to the table. With a foundation laid by Fast.ai's Deep Learning course, Radek has honed his skills through extensive project work, emerging as a champion in Kaggle competitions. Radek sheds light on his role at NVIDIA, sharing valuable insights gained from his time there and delving into topics like meta-learning and the path to mastery in machine learning.

📲 Radek Osmulski Socials:
LinkedIn: https://www.linkedin.com/in/radek-osmulski-6b935794/?originalSubdomain=au
Twitter: https://twitter.com/radekosmulski?lang=en

📲 Mark Moyou, PhD Socials:
LinkedIn: https://www.linkedin.com/in/markmoyou/
Twitter: https://twitter.com/MarkMoyou

📗 Chapters
0:00:00 Opening
0:01:45 NVIDIA
0:04:37 Kaggle Competitions
0:11:11 Fast AI Course
0:15:12 Learning AI
0:17:06 NVIDIA Job Title
0:18:40 Learning's at NVIDIA
0:21:47 Meta Learning
0:28:13 Becoming Great at ML
0:30:36 Building Career Outside US
0:33:56 Hustling
0:38:19 Code Cleaning Practices
0:41:58 Programming Skills
0:48:50 Learning with LLM
0:51:51 Product Mindset for Learning
0:54:10 Tweeting Advice
1:01:02 Radek's Career Optimization Function
1:06:36 Advice for Writers
1:21:36 Outside Work Interests
1:23:53 Book Recommendations
1:27:56 Advice for High School, College Student and Working Professional
1:29:36 Rapid Round

Harpreet brings a trove of knowledge and insights to the table. In this engaging conversation, we explore various facets of his expertise, from his role in developer relations to his deep understanding of Large Language Models (LLMs) and their evaluation. Harpreet provides invaluable insights into the intricacies of LLM evaluation, shedding light on methodologies, challenges, and best practices. His deep understanding of LLMs stems from years of hands-on experience and a commitment to pushing the boundaries of AI research and development.

📲 Harpreet Sahota Socials:
LinkedIn: https://www.linkedin.com/in/harpreetsahota204/?originalSubdomain=ca
Twitter: https://twitter.com/DataScienceHarp

📲 Mark Moyou, PhD Socials:
LinkedIn: https://www.linkedin.com/in/markmoyou/
Twitter: https://twitter.com/MarkMoyou

📗Chapters
00:00 Intro
02:10 What problems DECI AI is solving?
04:42 DeciLM-7B
06:41 Supergradients YOLO-NAS
09:03 Deci Team
09:40 DeciLM-7B vs Mistral 7B
14:54 LLM Evaluation
21:41 Fine tuning 7B and 13B Models
24:16 Fine tuning in RAG
29:24 Deploying Gen AI in Workloads
31:59 Non Technical Executives
33:15 Harpreet's Career Journey
38:34 Developer Relations Position
41:37 Your personal brand is an investment
48:20 Finding your own vibe
53:12 Data Science in 2024
56:32 Podcasting
01:05:02 When to quit
01:11:45 Career Optimization Function
01:18:03 Harpreet's Obsession besides Gen AI
01:19:30 AI for Babies
01:21:26 Curiosity
01:24:08 Changing Jobs
01:28:32 Book Recommendations
01:30:21 Advice for High Schooler, College Student and Professionals
01:32:15 Fav Meals
01:33:15 Legacy to leave behind.

Demetrios Brinkmann, the Founder and CEO of MLOps Community, as he shares his expertise on the dynamic landscape of Machine Learning Operations (MLOps) and the transformative impact of Large Language Models (LLMs). He also highlights the importance of authenticity and fun in community building, addresses common challenges in MLOps, and shares his experiences leading a technical community.

📲 Demetrios Brinkmann Socials:
LinkedIn: https://www.linkedin.com/in/dpbrinkm/
Twitter: https://twitter.com/Dpbrinkm

📲 Mark Moyou, PhD Socials:
LinkedIn: https://www.linkedin.com/in/markmoyou/
Twitter: https://twitter.com/MarkMoyou

📗Chapters
00:00 - Intro
02:56 - How has MLOps changed after the LLM explosion?
10:14 - Convergence of MLOps
13:37 - MLOps in Tech-First vs Legacy Companies
17:24 - Rise of MLOps (LLM) Companies
20:35 - Generative DevOps
22:08 - How Big is the MLOps Community?
24:10 - Inception of the MLOps Community
30:30 - Authenticity and Fun Are the Key
37:43 - Most Commonly Asked Problems/Topics in MLOps
41:28 - Learning New Things
47:00 - Leading a Technical Community
50:36 - Monetizing the Community
59:44 - Why isn't the MLOps Community on Discord?
01:02:48 - Importance of a Meme Channel
01:04:37 - Do Lurkers Add Value to the Community?
01:07:48 - Mistakes Made While Building the Community
01:14:41 - Lessons Learned on Selling
01:19:06 - Advice for Podcasters
01:26:58 - Motivation
01:28:30 - Success
01:34:12 - Book Recommendations
01:40:24 - Rapid Round

Sujatha Sundararaman, a distinguished engineering leader and Head of Search, Understanding, and Voice at Google's YouTube Music (YTM). With a stellar track record in both startup and enterprise environments, Sujatha excels in steering ideas from conception to successful, revenue-generating products. 

📲 Sujatha Sundararaman Socials:
LinkedIn: https://www.linkedin.com/in/sujatha-sundararaman-49a339/

📲 Mark Moyou, PhD Socials:
LinkedIn: https://www.linkedin.com/in/markmoyou/
Twitter: https://twitter.com/MarkMoyou

📗 Chapters
00:00 Intro
02:17 LLM in Music Recommendations
04:28 Why music is important?
08:05 Exploration Vs Exploitation
11:48 Prevalence of remix songs
15:07 Why music recommendation problem is difficult?
20:30 Large datasets and diversity of users
25:54 Handling User experience and expectations in music
28:20 Categorizing Music and Content Understanding
32:28 Glocal Music Behavior
36:20 Multilingual Users
41:22 Building trust with YT music
46:19 Generative AI in Music
58:54 How dance drives music listening behavior?
01:01:17 Startup Lessons
01:06:02 Career Optimization Function
01:08:20 Networking and Switching Companies
01:11:14 Mentors
01:15:46 Managing a Team
01:18:43 How to manage world class teams?
01:21:32 Advice for Investors in AI space
01:23:32 Advice for startup founders
01:25:35 What value MBA provides?
01:28:58 3 book recommendations
01:32:15 How to approach job market?
01:37:50 Rapid Round

Igor Jablokov, the visionary Founder and CEO of Pryon, an AI company at the forefront of innovation in enterprise knowledge management. In this captivating podcast episode, we delve into Igor's remarkable journey, from his formative years at IBM to his pivotal decision to establish Pryon. Fueled by a deep-seated impatience with the pace of AI commercialization, Igor embarked on a mission to bridge the gap between AI invention and real-world application.

📲 Igor Jablokov Socials:
LinkedIn: https://www.linkedin.com/in/ijablokov/
Twitter: https://twitter.com/ijablokov

📲 Mark Moyou, PhD Socials:
LinkedIn: https://www.linkedin.com/in/markmoyou/
Twitter: https://twitter.com/MarkMoyou

📗 Chapters
00:00 Intro
02:39 What problem are you solving at Pryon?
04:48 Navigating Enterprise LLM
06:59 Data Quality
09:17 Sustainable Compute
15:16 Responsive Companies
20:03 Capturing Legacy Information
23:04 Connectors
26:16 Scaling and Energy Efficiency
28:46 Content Understanding (Images and Video)
30:46 LLMs
35:49 Enterprise Knowledge Management
43:41 Search
47:42 Philosophy on Running Pryon
54:19 Attaining $100 million Funding
01:03:08 Attracting and Retaining Talent
01:10:55 Career Optimization Function

In this episode, Mike shares his extensive expertise in building and leading high-performance machine learning teams, delivering cutting-edge data products, and leveraging state-of-the-art technologies for various real-world applications. His contributions in text comprehension using large language models (LLMs), image recognition, Graph Neural Networks (GNNs), learned representation-based recommender systems, targeted advertising, time series forecasting, user understanding, and customer analytics are nothing short of awe-inspiring.

📲 Mike Tamir Socials:
LinkedIn: https://www.linkedin.com/in/miketamir/
Twitter: https://twitter.com/MikeTamir

📲 Mark Moyou, PhD Socials:
LinkedIn: https://www.linkedin.com/in/markmoyou/
Twitter: https://twitter.com/MarkMoyou

📗 Chapters
00:00 Inro
02:29 Shift in LLM Paradigm
08:50 First Impression of ChatGPT
15:00 LLM Fad
21:00 Mastering LLM 
26:48 Test & Evaluation
35:08 Hallucinations
37:40  Retrieval-Augmented Generation (RAG)
40:41 Large vs Small Model & Enterprise vs Open Source Models
52:44 LLM Strategy for Teams 
58:01 Evolution of Instructions tailored towards agent
59:52 Prediction Hypothesis  
01:01:52 Startup
01:09:50 Advice for VCs
01:13:12 Differences in Data Science Culture
01:16:38 Approach for Leading Teams
01:18:18 Reading
01:19:20 Succeeding as a Team
01:21:07 Small vs Big Teams
01:23:06 What makes you really good at your job?
01:25:52 PhD and Teaching
01:31:22 Research Papers
01:32:35 Career Optimization Function
01:34:45 Book Recommendations
01:37:15 Rapid Round

Jacopo Tagliabue is one of the most honest and talented open-source contributors in the Data Science community, previously focusing on NLP and Recommender Systems. Currently, he eats pizza, plays very occasional tennis, and is building a data pipeline company, Bauplan.

A true founder and builder at heart, Jacapo and team pioneered the movement for doing data science at reasonable scale. His NLP company Tooso, was acquired by Coveo where he led Data Science for search all the way to an IPO.

He 
- has been educated at top institutions such as UNISR, SFI, and MIT
- has collaborated on open-source research with Stanford, Netflix, Farfetch, and NVIDIA
- has invested in startups and funds as an LP 
- teaches machine learning at NYU as an adjunct professor

📲 Jacopo Tagliabue Socials:
LinkedIn: https://www.linkedin.com/in/jacopotagliabue/ (don't try to sell him anything)
Twitter: https://twitter.com/jacopotagliabue
Website: https://jacopotagliabue.it/

📲 Mark Moyou, PhD Socials:
LinkedIn: https://www.linkedin.com/in/markmoyou/
Twitter: https://twitter.com/MarkMoyou

📗 Chapters
00:00 Intro
03:34 LLM Progress
09:07 Small LLMs
11:38 Search and Information Retrieval
15:12 Recommender System
19:20 Robust Recommender System
22:37 Hardest Domain for Recommender System
24:54 Fake Datasets for Research
26:48 Mistakes in Building Recommender System 
30:10 Building a Reasonable Recommender System
33:46 Reasonable Scale
38:17 What Bauplan is trying to accomplish?
44:41 Data Processing
50:05 Build Vs Buy 
55:12 Building a New Startup
57:35 Tennis
58:31 Investments in ML Community
01:02:14 First Company
01:10:37 Advice for Founders/Lessons Learned from First Startup
01:15:03 PhD
01:18:39 Career Optimization Function
01:23:06 Advice for Open Source Contribution (Do the freaking commit!)
01:27:30 Journal for Failed Papers
01:29:53 Book Recommendations
01:34:00 Advice for High Schooler, College Student & Professional
01:36:10 Rapid Round