Shabaz Patel

Sr. AI Architect @Adobe | Prev. Director (Distinguished) AI Eng for AI Search and Discovery @BestBuy | @Stanford AI Lab, @oneconcerninc ML, @datmoAI Co-Founder

[Shabaz Patel @shabazbpatel]

Very excited that our book, Shipping Machine Learning Systems: A Practical Guide to Building, Deploying, and Scaling in Production, is now available.

Early in my career, Martin Kleppmann’s Designing Data-Intensive Applications gave me a framework for thinking in systems. It stuck with me not because of flashy techniques, but because it was clear, opinionated, and grounded in real-world constraints.

When I was shipping ML systems at my startup, Datmo, and later at One Concern and Best Buy, I kept running into challenges no book had fully prepared me for. My co-authors, Anand and Mohamed, saw the same patterns in their own work.

- Systems, models, and agents that silently decayed in production
- Debugging pipelines with logs, traces, and version control
- Evaluation setups misaligned with actual product outcomes
- CI frameworks using LLM-as-a-judge to catch regressions
- Architectures that didn’t survive first contact with users
- Constantly shifting constraints: latency, cost, throughput, org structure

These are the kinds of issues that led to this book.

We structured it around the crawl → walk → run progression most teams follow (or should). The goal was to document what we wish we’d known at each stage, and the tradeoffs that matter in practice.

Whether you’re working with traditional ML models or AI agents, building infra or running models in prod, I hope this book helps.

Thanks to many who have reviewed or contributed to the book! @chipro @Hassan_Sawaf, Riham S., @emadelwany, @_moelgendy, Ritesh Bajaj, @anders_ai, Sahil Khanna, @qasimmunye, @LawLM @qasim31wani @melissapan and others.

Cambridge: https://t.co/g5S588Gkqt
Amazon: https://t.co/3zi3Xub8vG

[Shabaz Patel @shabazbpatel]

Congratulations to the @hqfabi team on the launch! Step change for data analytics. Connect to different data sources, ask questions in plain english, and get clear, reliable answers. One can turn the analysis into dashboards/workflows in minutes. This is how AI should speed up decision-making.

[marc @marcfdupuis]

Today we’re introducing @hqfabi 2.0 - Your AI analyst for ALL your data

Connect to *any* data source.

Analyze it using an AI agent that understands (and learns!) your business.

Turn your insights into dashboards or workflows in minutes.

[Qasim Wani @qasim31wani]

Highly recommend folks transitioning to ML in-the-wild read this book!

Lots of insights shared by the team

[Melissa Pan @melissapan]

Very honoured to have🤖MAST��� (our paper: why do multi-agent LLM systems failed) featured in @shabazbpatel's new book!! 📖

Happy 2026 reading :D

[Jeff Dean @JeffDean]

Performance Hints

Over the years, my colleague Sanjay Ghemawat and I have done a fair bit of diving into performance tuning of various pieces of code. We wrote an internal Performance Hints document a couple of years ago as a way of identifying some general principles and we've recently published a version of it externally.

We'd love any feedback you might have!

Read the full doc at: https://t.co/jej95g236P

[Shabaz Patel @shabazbpatel]

https://t.co/FRPbcKjD2Y is the Cursor for data visualization and BI.

[marc @marcfdupuis]

📣 Today we’re officially announcing our launch of @hqfabi Workflows!

For those of your who follow us, you know that we believe in two things:

1️⃣ Finding insights from your data should be dead simple regardless of where your data lives. Some teams have everything flowing to awesome data warehouse ( @Snowflake, @Databricks, @MotherDuck, @ClickHouseDB to name a few), others live in Google Sheets or @Airtable. They each have their purpose, insights shouldn’t be limited to teams that just use one or the other.

2️⃣ Insights should be delivered where the team works. Dashboards are great - especially when they have AI natively embedded like ours do ;) - but rely on you remembering where to find said dashboard to use it. More impactful: Bring the insights to your users in email, @SlackHQ, or Google Sheets.

We’re bringing all these pieces together with Workflows.

→ Low-code Integration Cells: Using Python, the lingua franca of data analysis, you can push dataframes, images and variables to Google Sheets, Slack, or email in just a few clicks.

→ AI Analysis Cells: Want to have AI generate a summary to send to Slack? Or need the AI to process a dataframe column using a prompt? No problem! Our AI enrichment and summarization cells make that so simple even I can do it!

→ Workflow Canvas View: Get a birds eye view of your entire workflow to see how all the pieces work together. Tip: this is a GREAT way to quickly understand what one of your coworkers built when they share their work with you.

Much more to come on this front. I believe data teams will spend less time doing one-off analysis in the future and more time building “insight pipelines”. We’re building the foundation for this.

Read all about why we built Workflows, what features are in it, and what you can do with them in our latest blog 👉: https://t.co/mYyOrjEMyg

PS: We’d love your support on ProductHunt as well! https://t.co/8AKpNlusxh

[Anthropic @AnthropicAI]

THE WAY OF CODE, a project by @rickrubin in collaboration with Anthropic:

[Shabaz Patel @shabazbpatel]

@chipro Congrats Chip!

[Shabaz Patel @shabazbpatel]

@DGukesh Congratulations!

[Shabaz Patel @shabazbpatel]

I am excited to see new developments in ML/AI that can benefit from this kind of computing capability.

[Sundar Pichai @sundarpichai]

Introducing Willow, our new state-of-the-art quantum computing chip with a breakthrough that can reduce errors exponentially as we scale up using more qubits, cracking a 30-year challenge in the field. In benchmark tests, Willow solved a standard computation in <5 mins that would take a leading supercomputer over 10^25 years, far beyond the age of the universe(!).

[Shabaz Patel @shabazbpatel]

GenAI is set to revolutionize many industries including e-commerce search. It will be smarter, faster, and more intuitive. Finding the right product will become much easier.

[Shabaz Patel @shabazbpatel]

I am excited about a Black Friday release that helps users find the perfect tech gifts. Connect with me if this sounds exciting to you!

[Shabaz Patel @shabazbpatel]

I am excited about a Black Friday release that helps users find the perfect tech gifts. Connect with me if this sounds exciting to you!

[Shabaz Patel @shabazbpatel]

@SaMoRaAli @_MoustafaM This role is focused on a hybrid work model near the office locations.

[Shabaz Patel @shabazbpatel]

I am looking for AI/ML Engineers to build impactful models and AI agents for the e-commerce sector, serving millions of users. If this sounds interesting to you, please reach out!

https://t.co/17n20DxE21

[Shabaz Patel @shabazbpatel]

@finbarrtimbers More traffic onto meta AI answer engine products and thereby revenue growth and taking some Ads market share of Google.

[Shabaz Patel @shabazbpatel]

@virattt @cohere @weaviate_io How you are evaluating the financial RAG system?

[Shabaz Patel @shabazbpatel]

@soumithchintala @bhack_10 @JeffDean @SingularMattrix @fchollet @dwarak @GoogleDeepMind It would benefit the community to provide comparative studies of both products, highlighting their strengths and best use cases. A collaborative study on this is ideal.

[Shabaz Patel @shabazbpatel]

@RichardSocher @databricks @YouSearchEngine Great! How do you currently deploy your models? Are you using vLLM, Nvidia Triton, or a custom inference engine?

[Shabaz Patel @shabazbpatel]

@xai has released a Mixture-of-Experts model with 314B parameters where only 25% of weights are active on a given token.
https://t.co/erxd9qlLhn

[Shabaz Patel @shabazbpatel]

@eugeneyan Suggest you look into this tool: https://t.co/TNrmKC42Gv