From Scratch Podcast

a podcasting group focused on giving voices to great people at the intersection of engineering, research and startups. founded by @omeedtehrani and @laithaustin

[From Scratch Podcast @fspodofficial]

we have 1.3k views, almost 30 watch hours and 75+ subs within the span of 10 days. thank you so much everyone for supporting.

the episode with Gleb from Humanoid robotics should be out sometime next week!

[From Scratch Podcast @fspodofficial]

@jeff_weinstein @stripe cool idea

[omeed @omeedtehrani]

Meet Gleb Zarin. This is an old episode but is also posted on @YouTube! We are trying to be more active on @X with our content. Gleb is the lead of locomotion at Humanoid Robotics. Please check them out - they are revolutionizing robotics at @TheHumanoidAI. In this conversation, Gleb discusses the current state and future of humanoid robotics, addressing challenges such as funding, the impact of generative AI, and the importance of reinforcement learning. He emphasizes the significance of simulation and real-world testing, the role of human oversight, and the potential for robotics to transform industries. Gleb also shares insights on the collaborative nature of robotics development and the need for effective communication among teams. The discussion touches on the evolution of robotics, the challenges of deployment, and the future of human-robot interaction.

Chapters: 00:00:00 Introduction and Context Setting

00:00:18 Challenges in Humanoid Robotics

00:03:52 Funding and Investment in Robotics

00:06:29 Advancements in Locomotion and Reinforcement Learning

00:14:14 The Role of Simulation in Robotics

00:19:45 Machine Learning vs. Traditional Methods in Robotics

00:24:02 Future Directions in Robotics and Compute

00:26:27 The Role of Cloud in Robotics

00:27:36 Advancements in Simulation for Robotics

00:30:01 Testing Robots: Reality vs Simulation

00:31:31 Imitation Learning in Robotics

00:34:04 Building a Robot's First Step

00:35:37 Open Source Models in Robotics

00:38:14 Challenges in Robotics Engineering

00:40:50 Communication Challenges in Robotics Teams

00:42:01 Deployment Challenges in Robotics

00:42:28 Comparing Self-Driving and Robotics

00:44:43 Translating Self-Driving Experience to Robotics

00:47:44 Brittleness in Robotics

00:52:07 Simulating Variability in Robotics

00:52:50 Generative Models and Robotics

00:54:28 Acoustics and Robotics

00:57:05 The Humanoid Form Factor in Robotics

01:00:42 Designing Humanoid Robots: The Basics

01:01:47 Challenges in Humanoid Robotics

01:03:13 The Role of AI in Robotics

01:05:46 Opportunities in Humanoid Robotics

01:07:46 Navigating Regulations in Robotics

01:10:29 The Future of Human-Robot Interaction

01:14:33 Rapid Fire Questions: Insights and Opinions

[omeed @omeedtehrani]

New @fspodofficial episode! 🎙️ This is also posted on our YouTube!

Meet Areg Melik-Adamyan. Areg is the Head of Compiler Engineering for the MAX platform at @Modular (formerly a long-time leader of AI compiler efforts at Intel). His work spans MLIR enablement, Triton/XPU backends, and end-to-end AI software stacks—sitting right at the intersection of language design, runtime systems, and modern accelerators. In this episode, we dig into the new role of compilers in AI: why the current compute ecosystem is fragmented, how unification unlocks performance and portability, and what it takes to debug and optimize real LLM/AI workloads. We also talk Mojo’s ambitions, how foundational CS knowledge compounds for aspiring compiler engineers, and where Modular is headed as a universal compute layer.

Chapters:
00:00:00 Areg’s path from EDA to AI compilers
00:04:12 What a modern “AI compiler” actually does 00:08:45 Why the AI compute stack is fragmented (and how to unify it)
00:13:30 MLIR, Triton, and building for heterogeneous hardware
00:18:05 Debugging AI workloads vs. traditional systems 00:23:41 LLMs changing how we develop and optimize code
00:28:50 Mojo’s goals: performance, safety, and approachability
00:34:22 Lessons from Intel: hardware realities & software leverage
00:39:10 Foundational CS for future compiler engineers 00:44:37 MAX as a compute layer: portability without hand-tuning
00:50:15 The future: multi-target optimization & transparent acceleration
00:56:02 Advice for students and early-career builders 01:00:20 Closing thoughts: the next decade of AI compilers\

[omeed @omeedtehrani]

We’re back on @fspodofficial 🎙️ after a long break—and we’re kicking off the new run of episodes with Areg Melik-Adamyan on the From Scratch Podcast. Areg leads compiler engineering for Modular’s MAX platform, after 14+ years at Intel where he directed teams building next-gen AI compilers, enabled Triton on Intel XPU, and contributed upstream to MLIR.

What makes Areg stand out isn’t just his role today—it’s the breadth of his background: managing the GNU toolchain for Xeon/Atom, architecting cloud-native data platforms across Modin/Arrow/Ray/Pandas/PyTorch, launching NFF-Go for telco packet processing, filing multiple patents, and even teaching graduate-level courses in Computer Architecture and Distributed Systems. Few engineers see the stack as holistically—from silicon to compilers to real workloads—as Areg does.

Huge shoutout to @clattner_llvm (LLVM & Swift) and the team at @Modular for the vision behind Mojo and MAX. A unified, high-performance compute layer for AI isn’t just a nice-to-have—it’s how we make systems faster without heroics or vendor lock-in. Areg’s work is right at that inflection point.

🎧 We’re also back on a consistent cadence with a line of incredible guests already recorded (and more incoming).
🎥 Plus, we’re hiring creatives to level up production—editing, motion/graphics, thumbnails, and social distribution. If you obsess over polish, pacing, and storytelling, DM me.

🔗 Episode link: https://t.co/uTBNgP5vYZ

If you care about compilers, MLIR/Triton, or making LLMs faster without touching model weights—this one’s for you.

#ai #compilers #mlir #triton #mojo #systems #modular #fromscratchpodcast #performance #infrastructure #hiring #creators

[From Scratch Podcast @fspodofficial]

new podcast this week!

[omeed @omeedtehrani]

Check out the new episode with @mattklein123! We dug into @bitdriftio and @EnvoyProxy!

https://t.co/5RKRpF0uPf

[From Scratch Podcast @fspodofficial]

@elonmusk @joerogan us next?

[omeed @omeedtehrani]

Our podcast is growing! If you’re a startup founder, or someone building something that is changing the world… DM me. I’d love to have you on the podcast. You can also reach out: [email protected]

Think of it as a way for you to share your story with the world while marketing your company in its early stages as well. :)

[From Scratch Podcast @fspodofficial]

please make sure to follow all instructions to be entered: https://t.co/ElU9Ai1brI

[From Scratch Podcast @fspodofficial]

GIVEAWAY! We are doing a @nvidia jetson orin nano super computer developer kit giveaway! The rules are simple to enter:

(1) subscribe to our channel with notifications on.
(2) follow our hosts.
(3) like and repost to your network.
(4) comment how you’ll use it!

[From Scratch Podcast @fspodofficial]

@despxwned hi

[From Scratch Podcast @fspodofficial]

Our podcast is now available on Spotify. Are you interested in Humanoid robotics? Learn from an industry expert from @TheHumanoidAI. Available on YouTube as well!

https://t.co/jj7BFxS4Rr

[From Scratch Podcast @fspodofficial]

CHECK OUT EPISODE #2!

https://t.co/a4ivtJVaoZ

[From Scratch Podcast @fspodofficial]

EPISODE #2 WITH @zaringleb, LEAD OF LOCOMOTION AT HUMANOID ROBOTICS, IS OUT NOW!

https://t.co/fJ9pZ2xfoI

[From Scratch Podcast @fspodofficial]

EPISODE #2 WITH @zaringleb, LEAD OF LOCOMOTION AT HUMANOID ROBOTICS, IS OUT NOW!

https://t.co/fJ9pZ2xfoI

[From Scratch Podcast @fspodofficial]

What are some of the core differences between working on self-driving and robotics as an engineer? @zaringleb gives us an overview of some of the most important similarities and differences from his time at @wayve_ai and @TheHumanoidAI .

[From Scratch Podcast @fspodofficial]

@andrewchen hi

[From Scratch Podcast @fspodofficial]

episode #2 with one of the leads of locomotion at Humanoid should be coming out sometime soon as well!

[From Scratch Podcast @fspodofficial]

HIRING: we are looking for someone to help us on the social media / marketing side of the podcast, including outreach. Please DM us if you or anyone else is interested! #podcast #hiring