Agentic Conversations (formally mlops.community)

Latest Episodes:

Data Engineering for ML // Chad Sanderson // Coffee Sessions #117 Aug 19, 2022

Show notes

MLOps Coffee Sessions #117 with Chad Sanderson, Head of Product, Data Platform at Convoy, Data Engineering for ML, co-hosted by Josh Wills.

// Abstract
Data modeling is building relationships between core concepts within your data. The physical data model shows how the relationships manifest in your data environment, but then there's the semantic data model, the way that entity relationship design is extracted away from any data-centric implementation.
Let's have the good old fun of talking about why data modeling is so important!

// Bio
Chad Sanderson is the Product Lead for Convoy's Data Platform team, which includes the data warehouse, streaming, BI & visualization, experimentation, machine learning, and data discovery.
Chad has built everything from feature stores, experimentation platforms, metrics layers, streaming platforms, analytics tools, data discovery systems, and workflow development platforms. He’s implemented open source, SaaS products (early and late-stage) and has built cutting-edge technology from the ground up. Chad loves the data space, and if you're interested in chatting about it with him, don't hesitate to reach out.

// MLOps Jobs board

jobs.mlops.community

MLOps Swag/Merch
https://mlops-community.myshopify.com/

// Related Links
https://odsc.com/speakers/scaling-machine-learning-with-data-mesh/ https://docs.google.com/presentation/d/1rVtltHkRkP_JaGZdkAS3U_SXfr5Gg-RP980FKXh0YNU/edit?usp=sharing
Josh Wills will be teaching a course on Data Engineering for Machine Learning in September here:
https://www.getsphere.com/ml-engineering/data-engineering-for-machine-learning

--------------- ✌️Connect With Us ✌️ -------------
Join our Slack community: https://go.mlops.community/slack
Follow us on Twitter: @mlopscommunity
Sign up for the next meetup: https://go.mlops.community/register
Catch all episodes, blogs, newsletters, and more: https://mlops.community/

Connect with Demetrios on LinkedIn: https://www.linkedin.com/in/dpbrinkm/
Connect with Josh on LinkedIn: https://www.linkedin.com/in/josh-wills-13882b/
Connect with Chad on LinkedIn: https://www.linkedin.com/in/chad-sanderson/

Timestamps:
[00:00] Introduction of the new co-host, Josh Wills
[00:54] Introduction to Chad Sanderson
[01:46] Josh will lead a course on Machine Learning in mid-September
[02:16] Data modeling blog post of Chad
[06:10] Idea of Strategy
[09:40] Modern cloud data warehouses
[17:01] Layering on contracts
[20:38] Scaling at larger companies
[25:30] Carrot-stick strategy
[34:27] Second and third-order effects
[39:53] Stockholm Syndrome
[41:22] Quality checks at Slack
[45:28] Success in two main ways, according to Chad
[47:35] Completely and utterly different universes
[53:42] Product use case to push semantic events
[56:00] Pattern of analysis of the sequence of events
[57:23] Wrap up

Scaling Machine Learning with Data Mesh // Shawn Kyzer // Coffee Sessions #116 Aug 17, 2022

Show notes

MLOps Coffee Sessions #116 with Shawn Kyzer, Principal Data Engineer at Thoughtworks (Spain), Scaling Machine Learning with Data Mesh, co-hosted by Adam Sroka.

// Abstract
You can't just get something done by using tools. You need to go much deeper than that, and it is very clear that Data Mesh is the same thing. You have to educate the organization about the movement.
In this session, Shawn broke down the cultural piece of data mesh and how many parallels there are with the MLOps Movement when it comes to the cultural side of MLOps.

// Bio
Shawn is passionate about harnessing the power of data strategy, engineering, and analytics in order to help businesses uncover new opportunities. As an innovative technologist with over 13 years of experience, Shawn removes technology as a barrier and broadens the art of the possible for business and product leaders. His holistic view of technology and emphasis on developing and motivating strong engineering talent, with a focus on delivering outcomes whilst minimising outputs, is one of the characteristics that sets him apart from the crowd.

Shawn’s deep technical knowledge includes distributed computing, cloud architecture, data science, machine learning, and engineering analytics platforms. He has years of experience working as a consultant practitioner for a variety of prestigious clients, ranging from secret clearance-level government organizations to Fortune 500 companies.

// MLOps Jobs board
jobs.mlops.community

MLOps Swag/Merch
https://mlops-community.myshopify.com/

Connect with Demetrios on LinkedIn: https://www.linkedin.com/in/dpbrinkm/
Connect with Adam on LinkedIn: https://www.linkedin.com/in/aesroka/
Connect with Shawn on LinkedIn: https://www.linkedin.com/in/shawn-kyzer-msit-mba-b5b8a4b/

Timestamps:
[00:00] Introduction to Shawn Kyzer
[00:43] Takeaways
[04:00] Data Mesh for ML projects
[11:22] The signal for the exploratory part of a new modeling project
[14:13] Ownership and centralization
[16:20] Lack of technology and some implementations of literature
[17:10] Python stronghold from Microsoft blogs
[23:09] Integration with self-serve data platform
[25:31] Starting a platform team
[30:04] Quick wins
[32:09] Metrics monitoring
[34:18] Metrics break up
[38:32] Limit to capabilities and not worth doing
[41:39] Culture and technology hold
[44:03] Setting the foundation
[46:53] Unforeseen benefits
[52:19] Lightning question

How Hera is an Enabler of MLOps Integrations // Flaviu Vadan // Coffee Sessions #115 Aug 14, 2022

Show notes

MLOps Coffee Sessions #115 with Flaviu Vadan, Senior Software Engineer at Dyno Therapeutics, How Hera is an Enabler of MLOps Integrations, co-hosted by Vishnu Rachakonda.

// Abstract
Flaviu talks about the internal ML platform at Dyno Therapeutics called Hera. His team uses Hera as an internal innovation engine to help discover new breakthroughs with machine learning in the biotech healthcare industry.

// Bio
Flaviu is a Senior Software Engineer at Dyno Therapeutics, the leading organization in the design of novel gene therapy vectors with transformative delivery properties for a vast landscape of human diseases. Flaviu comes from a background focused on Bioinformatics, which is a field that combines Computer Science, Mathematics, and Biology. He took stints in academia by working as a research assistant in Computer Science and Bioinformatics labs before joining Dyno Therapeutics to work on machine-guided design of adeno-associated viruses (AAVs).

At Dyno, Flaviu works on compute and core infrastructure, DevOps, MLOps, and approaches that combine AI/ML to design AAVs in silico. He is also the author and maintainer of Hera, a Python SDK that facilitates access to Argo Workflows by making workflow construction and submission easy and accessible.

// MLOps Jobs board
jobs.mlops.community

MLOps Swag/Merch
https://mlops-community.myshopify.com/

// Related Links

Connect with Demetrios on LinkedIn: https://www.linkedin.com/in/dpbrinkm/
Connect with Vishu on LinkedIn: https://www.linkedin.com/in/vrachakonda/
Connect with Flaviu on LinkedIn: https://www.linkedin.com/in/flaviuvadan/

Timestamps:
[00:00] Introduction to Flaviu Vadan
[00:50] Takeaways
[02:06] Share this episode with a friend!
[03:20] What Dyno does
[05:44] CRISPR and Gene Editing
[06:21] Kidney transplants and using pig organs
[07:31] Deciding what genes to put in the body
[07:48] Role of ML at Dyno
[10:07] Higher dose
[13:41] Process of Machine Learning Deployment and Productionizing at Dyno
[16:22] Proliferation of models
[17:31] Building the internal platform
[19:37] Interaction with data, translation to compute layer, evaluation
[24:21] Venn diagram for MLOps
[27:06] Leveraging Argo Workflows
[30:34] Hera
[35:28] Open sourcing
[38:44] Human power at Dyno
[41:17] Wrap up

Product Enrichment and Recommender Systems // Marc Lindner and Amr Mashlah // Coffee Sessions #114 Aug 10, 2022

Show notes

MLOps Coffee Sessions #114 with Marc Lindner, Co-Founder, COO, and Amr Mashlah, Head of Data Science of eezylife Inc., Product Enrichment and Recommender Systems, co-hosted by Skylar Payne.

// Abstract
The difficulties of making multi-modal recommender systems. How can it be easy to know something about a user but very hard to know the same thing about a product and vice versa? For example, you can clearly know that a user wants an intellectual movie, but it is hard to accurately classify a movie as intellectual and fully automated.

// Bio
Marc Lindner has a background in Knowledge Engineering. He's always extremely product-focused with anything to do with Machine Learning.
Marc built several products working together with companies such as Lithium Technologies, etc., and then co-founded eezy.

Amr Mashlah
Amr is the head of data science at eezy, where he leads the development of their recommender engine. Amr has a master's degree in AI and has been working with startups for 6 years now.

// MLOps Jobs board

jobs.mlops.community

MLOps Swag/Merch
https://mlops-community.myshopify.com/

// Related Links
Children of Time book by Adrian Tchaikovsky:
https://www.amazon.com/Children-Time-Adrian-Tchaikovsky/dp/0316452505

Connect with Demetrios on LinkedIn: https://www.linkedin.com/in/dpbrinkm/
Connect with Skylar on LinkedIn: https://www.linkedin.com/in/skylar-payne-766a1988/
Connect with Marc on LinkedIn: https://www.linkedin.com/in/marc-lindner-883a0883/
Connect with Amr on LinkedIn: https://www.linkedin.com/in/mashlah/

Timestamps:

[00:00] Takeaways

[04:20] Introduction to Marc Lindner and Amr Mashlah

[06:08] eezylife Inc.

[08:43] Integration with different tools

[10:13] Richer data for eezy

[12:58] Challenges with different providers

[15:25] Labeling solutions

[16:00] Maodal.com

[18:18] Handling multi-modal recommendation

[20:11] Figuring out the process of ingesting the data

[23:16] Ontology of sorts

[27:21] Talking in-depth about technical pieces

[28:40] Handling cold start

[31:42] Bad recommendations

[37:24] Curation vs hard-coded rules

[39:53] Social features

[41:57] eezy's vision

[45:12] Suggested recommendations

[46:48] Preferences from inferred interactions

[51:45] Ethical considerations

[54:19] Wrap up

Building Better Data Teams // Leanne Fitzpatrick // Coffee Sessions #113 Aug 06, 2022

Show notes

MLOps Coffee Sessions #113 with Leanne Fitzpatrick, Director of Data Science of Financial Times, Building Better Data Teams, co-hosted by Mihail Eric.

// Abstract
We spent a lot of time talking about data tooling, but we may not have spent as much time talking about data organizations and efficiently running and organizing data teams.

What about starting with limitations instead of aspirations? Right constraints instead of the north star? In this session, let's learn more about a realistic take on the state of data organizations of today.

// Bio
Leanne is Director of Data Science at the Financial Times and is a passionate data leader with experience building and developing empowered data science and analytics teams in a variety of businesses. Leanne is in her element when developing and implementing strategic, technical, and cultural solutions to get machine learning and data science into the operational ecosystem.
Leanne is an active part of the data and technology community, sharing innovation and insights to encourage best practices, from Manchester, UK, to Austin, TX, and is an Advisory Panel Board Member. Outside of all things data, you can ask Leanne about her golf swing (it’s not good - yet), her passion for American Football (specifically the Cincinnati Bengals), her latest sewing project, and her love for good music, food, and whisky.

// MLOps Jobs board
jobs.mlops.community

MLOps Swag/Merch
https://mlops-community.myshopify.com/

// Related
Links Children of Time book by Adrian Tchaikovsky:
https://www.amazon.com/Children-Time-Adrian-Tchaikovsky/dp/0316452505

Connect with Demetrios on LinkedIn: https://www.linkedin.com/in/dpbrinkm/
Connect with Mihail on LinkedIn: https://www.linkedin.com/in/mihaileric/
Connect with Leanne on LinkedIn: https://www.linkedin.com/in/leanne-kim-fitzpatrick-29204341/

Timestamps:
[00:00] Introduction to Leanne Fitzpatrick
[04:23] Write us your suggestions!
[05:43] Tri-pawed dog called Seaweed!
[08:43] How to architect data teams
[14:44] Organizational deficiencies
[19:19] Tensions and conflicts for starters
[24:07] Misunderstandings from marketing
[25:59] The Middle Layer
[28:48] Data science work at publications
[31:11] Mystique of going to real-time
[35:29] Third parties with fraud
[37:40] Augmenting data practitioners with third-party tools
[41:00] Principle of reinventing the wheel and avoiding undifferentiated heavy lifting
[46:29] Different Abstraction Layer recommendations
[48:42] RN Production
[51:56] Will Python eat RN Production away?
[56:05] Julia as a dark horse
[56:39] Future of RN Production
[58:00] Rapid-fire questions

MLX: Opinionated ML Pipelines in MLflow // Xiangrui Meng // Coffee Sessions #112 Aug 03, 2022

Show notes

MLOps Coffee Sessions #112 with Xiangrui Meng, Principal Software Engineer of Databricks, MLX: Opinionated ML Pipelines in MLflow, co-hosted by Vishnu Rachakonda.

// Abstract
MLX is to enable data scientists to stay mostly within their comfort zone, utilizing their expert knowledge while following the best practices in ML development and delivering production-ready ML projects, with little help from production engineers and DevOps.

// Bio
Xiangrui Meng is a Principal Software Engineer at Databricks and an Apache Spark PMC member. His main interests center around simplifying the end-to-end user experience of building machine learning applications, from algorithms to platforms and to operations.

// MLOps Jobs board

jobs.mlops.community
MLOps Swag/Merch
https://mlops-community.myshopify.com/

// Related Links
Good Strategy Bad Strategy: The Difference and Why It Matters, book by Richard Rumelt:
https://www.amazon.com/Good-Strategy-Bad-Difference-Matters/dp/0307886239

Connect with Demetrios on LinkedIn: https://www.linkedin.com/in/dpbrinkm/
Connect with Vishnu on LinkedIn: https://www.linkedin.com/in/vrachakonda/
Connect with Xiangrui on LinkedIn: https://www.linkedin.com/in/mengxr/

Timestamps:
[00:00] Introduction to Xiangrui Meng
[00:39] Takeaways
[02:09] Xiangrui's background
[03:38] What kept Xiangrui in Databricks
[07:33] What needs to be done to get there
[09:20] Machine Learning passion of Xiangrui
[11:52] Changes in building that keep you fresh for the future
[14:35] Evolution core challenges to real-time and use cases in real-time
[17:33] DevOps + DataOps + ModelOps = MLOps
[19:21] MLFlow Support
[21:37] Notebooks to production debates
[25:42] Companies tackling Notebooks for production
[27:40] MLOoops stories
[31:03] Opinionated MLOps productionizing in a good way
[40:23] Xiangrui's MLOps Vision
[44:47] Lightning round
[48:45] Wrap up

More than a Cache: Turning Redis into a Composable, ML Data Platform // Samuel Partee // Coffee Sessions #111 Jul 30, 2022

Show notes

MLOps Coffee Sessions #111 with Samuel Partee, Principal Applied AI Engineer of Redis, More than a Cache: Turning Redis into a Composable, ML Data Platform, co-hosted by Mihail Eric. This episode is sponsored by Redis.

// Abstract
Pushing forward the Redis platform to be more than just the web-serving cache that we've known it to be up to now. It seems like a natural progression for the platform. We see how they're evolving to be this AI-focused, AI native serving platform that does vector similarity, feature storage, and provides those kinds of functionalities.

// Bio
A Principal Applied AI Engineer at Redis, Sam helps guide the development and direction of Redis as an online feature store and vector database.
Sam's background is in high-performance computing, including ML-related topics such as distributed training, hyperparameter optimization, and scalable inference.

// MLOps Jobs board
jobs.mlops.community

MLOps Swag/Merch
https://mlops-community.myshopify.com/

// Related Links
https://partee.io
Redis VSS demo: https://github.com/Spartee/redis-vector-search
Redis Stack: https://redis.io/docs/stack/
Github - https://github.com/Spartee
OSS org Sam co-founded at HPE/Cray - https://github.com/CrayLabs
This paper last year was some of the best research and collaborations Sam has been a part of. The Paper is published here: https://www.sciencedirect.com/science/article/pii/S1877750322001065?via%3Dihub
Do you really need an extra database for vectors? https://databricks.com/dataaisummit/session/emerging-data-architectures-approaches-real-time-ai-using-redis
Blink: The Power of Thinking Without Thinking by Malcolm Gladwell, Barry Fox, Irina Henegar (Translator): https://www.goodreads.com/book/show/40102.Blink

Connect with Demetrios on LinkedIn: https://www.linkedin.com/in/dpbrinkm/
Connect with Mihail on LinkedIn: https://www.linkedin.com/in/mihaileric/
Connect with Sam on LinkedIn: www.linkedin.com/in/sam-partee-b04a1710a

Timestamps:
[00:00] Introduction to Samuel Partee
[00:24] Takeaways
[02:46] Updates on the Community
[05:17] Start of Redis
[08:10] Vision for Vector Search
[11:05] Changing the narrative, going from the "Cache" for all servers and web endpoints
[14:35] Clear value prop on demos
[20:17] Vector Database
[26:26] Features with benefits
[28:41] AWS Spend
[30:39] Vector Database upsell model and bureaucratic convenience
[32:08] Distributed training hyperparameter optimization and scalable inference
[35:03] Core infrastructural advancement
[36:55] Tools movement to help
[39:00] Using Machine Learning at scale in numerical simulations with SmartSim: An application to ocean climate modeling (published paper)

[42:52] Future applications of tech to get excited about
[44:20] Lightning round
[47:48] Wrap up

Just Fetch the Data and then... // David Bayliss // Coffee Sessions #110 Jul 29, 2022

Show notes

MLOps Coffee Sessions #110 with David Bayliss, Chief Data Scientist of LexisNexis Risk Solutions, Just Fetch the Data and then... co-hosted by Vishnu Rachakonda.

// Abstract
Composing data to extract features can be a significant problem. Key factors are the data size, compliance restrictions, and real-time data. Ethics (and law) can drive extremely complex audit requirements. In the cloud, you can do anything - at a price.

// Bio
One of the creators of the world's first big data platform (HPCC), David has been tackling big data problems for two decades. A mathematician, compiler writer, and data sponge with more than five dozen patents spanning platforms, linking, and search.

Most inventors think outside the box; David can't even remember where the box is. He leads the team that creates their core Data Science methods used by hundreds of data scientists.

// MLOps Jobs board

MLOps Swag/Merch
https://mlops-community.myshopify.com/

Connect with Demetrios on LinkedIn: https://www.linkedin.com/in/dpbrinkm/
Connect with Vishnu on LinkedIn: https://www.linkedin.com/in/vrachakonda/
Connect with David on LinkedIn: https://www.linkedin.com/in/david-bayliss-426556a/

Timestamps:
[00:00] Introduction to David Bayliss
[01:03] Takeaways
[04:56] LexisNexis and David's role
[07:15] Evolution of LexisNexis in 20 years with so many use cases
[08:51] Role of David in structuring data for working with data change
[14:32] Data management and data access
[17:45] Unique challenges of scale, use case, and diversity at LexisNexis
[24:47] Tardis Iron Box
[30:05] Iron Box translation
[32:56] JVM for data science
[34:24] Iron Box meaning
[36:52] Metadata with PII
[39:08] Detrimental privacy / Hairy Kneecap Theory
[40:57] Speeding things up and Anonymized linking
[46:47] What kept David working at LexisNexis?
[50:30] Wrap up

Why You Need More Than Airflow // Ketan Umare // Coffee Sessions #109 Jul 23, 2022

Show notes

MLOps Coffee Sessions #109 with Ketan Umare, Co-founder and CEO of Union.ai, Why You Need More Than Airflow, co-hosted by George Pearse.

// Abstract
Airflow is a beloved tool by data engineers and Machine Learning Engineers alike. But when doing ML, what are the shortcomings, and why is an orchestration tool like that not always the best developer experience? In this episode, we break down what some key drivers are for using an ML-specific orchestration tool.

// Bio
Ketan Umare is the CEO and co-founder at Union.ai. Previously, he had multiple Senior roles at Lyft, Oracle, and Amazon, ranging from Cloud, distributed storage, Mapping (map-making), and machine-learning systems. He is passionate about building software that makes engineers' lives easier and provides simplified access to large-scale systems. Besides software, he is a proud father and husband, and enjoys traveling and outdoor activities.

// MLOps Jobs board
jobs.mlops.community

MLOps Swag/Merch
https://mlops-community.myshopify.com/

// Related Links
Zero to One: Notes on Startups, or How to Build the Future, Hardcover by Peter Thiel and Blake Masters:
https://www.amazon.com/Zero-One-Notes-Startups-Future/dp/0804139296

Connect with Demetrios on LinkedIn: https://www.linkedin.com/in/dpbrinkm/
Connect with George on LinkedIn: https://www.linkedin.com/in/george-pearse-b7a76a157/?originalSubdomain=uk
Connect with Ketan on LinkedIn: https://www.linkedin.com/in/ketanumare/

Timestamps:

[00:00] Introduction to Ketan Umare

[01:08] Takeaways

[04:03] Changes with Ketan after 2 years

[10:06] Diametrically opposed software and data infrastructure

[13:06] Airflow shortcomings

[21:14] Core abstractions

[24:56] Machine Learning specific packages

[28:25] Core target users

[30:50] Airflow foundational level

[33:46] Better visualization of data science pipelines

[36:39] Flyte for RL Deployment

[38:36] Role of signals in the system

[40:59] Everything in Flyte is run through an API

[42:20] Manual intervention in the system

[45:11] Union vs Flyte

[50:16] Security precautions

[51:56] Role of Union cloud

[56:20] Distinctiveness of Flyte pipelines

[01:01:11] Ketan in his current company

[01:02:25] Learning from Airflow

[01:04:58] Lightning round with Ketan

[01:09:59] Wrap up

ML Flow vs Kubeflow 2022 // Byron Allen // Coffee Sessions #108 Jul 19, 2022

Show notes

MLOps Coffee Sessions #108 with Byron Allen, AI & ML Practice Lead at Contino, ML Flow vs Kubeflow 2022 co-hosted by George Pearse.

// Abstract

The amazing Byron Allen talks to us about why MLflow and Kubeflow are not playing the same game!
ML flow vs Kubeflow is more like comparing apples to oranges, or as he likes to make the analogy, they are both cheese, but one is an all-rounder and the other a high-class delicacy. This can be quite deceiving when analyzing the two. We do a deep dive into the functionalities of both and the pros/cons they have to offer.

// Bio

Byron wears several hats. AI & ML practice lead, solutions architect, ML engineer, data engineer, data scientist, Google Cloud Authorized Trainer, and scrum master. He has a track record of successfully advising on and delivering data science platforms and projects. Byron has a mix of technical capability, business acumen, and communication skills that make me an effective leader, team player, and technology advocate.

See Byron write at https://medium.com/@byron.allen

// MLOps Jobs board

jobs.mlops.community

MLOps Swag/Merch
https://mlops-community.myshopify.com/

// Related Links

Connect with Demetrios on LinkedIn: https://www.linkedin.com/in/dpbrinkm/
Connect with George on LinkedIn: https://www.linkedin.com/in/george-pearse-b7a76a157/?originalSubdomain=uk
Connect with Byron on LinkedIn: https://www.linkedin.com/in/byronaallen/

Timestamps:
[00:00] Introduction to Byron Allen
[01:10] Introduction to the new co-host, George Pearse
[01:41] ML Flow vs Kubeflow
[05:40] George's take on ML Flow and Kubeflow
[07:28] Writing in YAML
[09:47] Developer experience
[13:38] Changes in ML Flow and Kubeflow
[17:58] Messing around ML Flow Serving
[20:00] A taste of Kubeflow through K-Serve
[23:18] Managed service of Kubeflow
[25:15] How George used Kubeflow
[27:45] Getting the Managed Service
[31:30] Getting Authentication
[32:41] ML Flow docs vs Kubeflow docs
[36:59] Kubeflow community incentives
[42:25] MLOps Search term
[42:52] Organizational problem
[43:50] Final thoughts on ML Flow and Kubeflow
[49:19] Bonus [49:35] Entity-Centric Modeling
[52:11] Semantic Layer options
[57:27] Semantic Layer with Machine Learning
[58:40] Satellite Infra Images demo
[1:00:49] Motivation to move away from SQL
[1:03:00] Managing SQL
[1:05:24] Wrap up

Our TOPPODCAST Picks

Follow Us

Stay Connected

Agentic Conversations (formally mlops.community)

Related Podcasts

Links

Stay Connected