Adi Polak
@AdiPolak
Director @ Confluent • Author • Databricks MVP • From distributed data to real-world AI systems
I earned a master's degree following my thesis in machine learning and can tell you right now that some of the BEST coders I've worked with never had formal education. It 💯% helped me land a job interview, but it does not make me a better coder/data scientist than anyone else.
LLMs aren’t the product — the system is. You need APIs, streaming or batch inference, monitoring, and scale. Training is step one. Shipping is the real job.
You don’t have to like it — but if you’re serious about AI engineering, you should at least understand what is required of you. Form your own opinion. Don’t outsource your thinking to clickbait headlines and conference hype.
AI tools are a great companion. Yet you still need to carefully review the outcome.
I'm a bit horrified that I've now memorized the git reset --soft HEAD~1 command...
You'd think the highest-paid sw engineers spend most of their time writing code. If you observe them, you'll find they read a lot of code and documents, talk with lots of people, and think a lot. Then they do what helps the business the most. THIS is how you win the AI…
If you are a heavy user of AI IDE's. You should know this. Productivity, doesn't always mean coding. It means listening to other people. Asking for help. Letting your mind take a break so that when you come back from examining the generated code, you can see it from a new…
Curious what you’ll think
You might already be skilled in ETL scheduling, analytics queries, or machine learning integration—but are you ready to support AI agents? In her 🆕 article in @TheNewStack, @AdiPolak dives into the skills data engineers need in the age of AI! ⤵️ thenewstack.io/data-engineeri…
RAG systems are like building a house: You need a solid foundation (retrieval infrastructure), strong walls (embedding pipelines), and a well-organized knowledge base (not a messy vector store). Hallucinations happen when the structure is weak.
Agents deleting production data? Whoop! Vibes coding started ok. Agents started ok too !
Heya all 👋 It's been long overdue! What’s up? What are you working on?
Statistics play an important role in genetics. For instance, statistics prove that numbers of offspring is an inherited trait. If your parent didn't have any kids, odds are you won't either.
Which tech company do you think will continue to exist 20 years from now? I’ll start, NVIDIA.
If you’re an AI Engineer and don’t understand event-driven systems, feature stores, vector DBs, or model pipelines — you’re just fine-tuning in the dark. AI isn’t just about models. It’s about building systems that work. Learn that, or get left behind.
If you’re an AI Engineer and don’t understand these yet, you’re just fine-tuning in the dark: • Event-driven systems • Feature stores • Vector DBs • Model pipelines AI isn’t just models. It’s systems. Learn to build them.
You don’t need a PhD to break into AI. Start here: •Learn Python •Build with OpenAI & Hugging Face APIs •Understand embeddings & prompts •Fine-tune a model on Colab •Ship something with FastAPI or Streamlit Projects > degrees.
The mindful AI Engineer: You never fail. If it doesn’t work the first time, just call it v1 of the pipeline. Debug, iterate, (scream), deploy.
AI Engineering = when a Data Scientist and a Data Engineer have a systems-savvy baby.