Linda Vivah

Linda Vivah Al/ML explained so it sticks⚡️
Al Engineer | Al Infra | 10+ yrs in tech | ex-AWS ☁️
singing-as-a-service 🎶 I mom x3 👶| 📍NYC

✨AI infrastructure✨ is a genuinely broad term. Here’s a mind map of what it actually is layer by layer🔹 Power & data cen...
09/03/2026

✨AI infrastructure✨ is a genuinely broad term. Here’s a mind map of what it actually is layer by layer

🔹 Power & data centers:
one modern GPU rack (NVIDIA’s GB200 NVL72) draws about 120 kW, roughly 10x a traditional server rack. Power is the fixed budget, so every layer above this one is partly about getting more compute out of the same watts

🔹 Compute:
NVIDIA did $89B in data center revenue last quarter. The hyperscalers now also build their own chips (TPU, Trainium, Maia, MTIA)

🔹 Networking:
thousands of chips sync up after every training step (sharing what they each just learned) and everything stalls until the slowest link finishes

🔹 Data & storage:
GPUs regularly sit below 50 percent utilization just waiting on data. People blame the GPUs when it’s usually the storage

🔹 Orchestration:
the layer that queues your job, starts it on every GPU at once and restarts it when one dies. Slurm, Kubernetes, Ray

🔹 Training & inference:
you train a model once but you pay to run it every day after. Inference (actually running the model for users) is now about two-thirds of all AI compute (Deloitte 2026)

🔹 Software:
CUDA, ROCm, XLA, the kernel and compiler layer that decides how much performance you actually get from the hardware

🔹 Observability & governance:
the dashboards can show healthy while the model is confidently wrong

Most of these layers already power traditional software….AI mainly moves the bottleneck: from database queries and web traffic to memory bandwidth, interconnect speeds & kilowatts.

✨ Save this for the next time you need to explain what AI Infrastructure is

08/30/2026

You can now add multiple Gmail accounts to ChatGPT 📩📩

go to ChatGPT > plugins > gmail > connect another account + you can nickname them ✨

Have been waiting for this to be available natively as I know many others

08/28/2026

Andrew Ng & the DeepLearning.AI team analyzed 10,000+ job postings + expert interviews & just shared the skills that matter most in AI engineering right now for developers 📚

The 4 skills:

🔹 Building & deploying AI applications:
AI systems have unpredictable outputs, so the work is learning the building blocks (RAG, evals, agentic workflows, context engineering) and using them to measure, steer, and govern how the system behaves

🔹 Software engineering fundamentals:
so much of software engineering is actually trade-offs (cost, speed, scalability, reliability, security). It’s making decisions, and if you don’t know the trade-offs, it’s hard to make those decisions

🔹 Using coding agents:
building a mental model of what agents can and can’t do, managing context, balancing planning vs ex*****on, and keeping current as the tools evolve

🔹 Shaping the build:
participating in the spec decisions themselves, which takes product sense and an understanding of business context and customer goals. Part of the skill is knowing when to spin up an MVP and when to build carefully

Comment SKILLS and I’ll send you the link to the report 📩

6ish weeks ago I sat for the industry’s first Context Engineer certification in beta at Databricks Data + AI Summit. The...
08/03/2026

6ish weeks ago I sat for the industry’s first Context Engineer certification in beta at Databricks Data + AI Summit. The results are in…

Excited to share that I am certified! ✅ & that the cert is now open to everyone!

💡 TLDR: The beta exam was only available at summit & it’s now open to everyone, online or in person at testing centers with the first sitting July 29.

A few takeaways if you’re thinking about taking it:

🧩 It’s scenario based. You’re given a situation and asked to pick the right approach, so it really tests how you think through designing agent systems, not what you memorized.

📖 Use the official guides. The exam was very aligned to the exam guide, so I’d get comfortable with every concept in there. There’s also an official AI prep guide you can hand to your AI for a study plan (super cool!)

🔁 The concepts transfer. It’s a Databricks cert, but the vocabulary is the field’s. If you’re building agent systems anywhere this is a great guide for context engineering in general.

Comment ✨context✨ & I’ll DM you the link that has the exam guide + AI prep guide 💬

As always, Happy Building! 😊

07/29/2026

Are you hearing the term ✨graph engineering✨ yet?

About 10 days ago, Peter Steinberger (creator of OpenClaw) tweeted this. I find it fascinating, and not because of the term itself. I look at everything through how people share information, & coining terminology is how tech has always worked: we align on a word so we can share and build the same thing together. What’s changed is the pace. The align-and-adapt cycle that used to take a field years is now happening in weeks.

What the terms actually mean:

🔹 Loop engineering (named by Addy Osmani back in June): one agent cycling act ➡️ observe ➡️ adjust until the job is done
🔹 Graph engineering: several of those looping agents wired together as nodes, with the graph deciding how work routes between them
🔹 The nuance: each node keeps its inner loop, and the graph itself can loop too

And the concepts here aren’t new (LangChain has been calling LangGraph work “graph engineering” for years, Anthropic mapped these patterns in 2024, Airflow has been a graph of tasks for a decade). What’s actually happening is bigger than everyone adopting a word at once: it’s decades of transferable systems thinking (graphs, orchestration, architecture) being applied to the newest layer we’re building on. Loops, nodes, and edges came from classic computer science; agents are just the newest thing we’re wiring with them.

I know a lot of people are tired of the terminology changing every month. But we’ve always renamed things as understanding evolved; AI just compressed the cycle. So much of getting the most out of these systems is how you think about them, and the vocabulary is the industry thinking out loud, faster than ever.

[ai engineering, agentic ai, system design]

07/16/2026

⚡ Lakebase vs Lakehouse in < 90 seconds. Why do you need both?

Michael Armbrust, Distinguished Engineer at Databricks , is back for another tech walk!

💡 TLDR: Lakehouse solved analytics. Lakebase does the same thing for the transactional side. Two workloads, one open format.

Back in the day, analytics over massive data meant a proprietary engine and a proprietary format. You loaded your data in and you were trapped. Lakehouse changed that: separate compute and storage, keep the data in an open format, and any engine can read it.

But analytics is only one kind of workload. The apps behind things like RAG don’t run huge queries. They run tiny operations that need answers in milliseconds. That’s a latency problem, and it’s a different type of data.

The interesting part: Lakebase applies the same open-format unlock to OLTP that Lakehouse applied to analytics, with open source Postgres underneath.

💻 Want to get hands-on? Everything here is testable on Databricks Free Edition

Comment ✨LAKE✨ & ill DM you the link

07/14/2026

The 4 main types of Agent Memory, explained in 40 seconds by Richmond Alake (Director of AI DevEx at Oracle) ⚡️

🧠 An agent’s context window is temporary by design. Without a persistence layer, everything it gathered during a session is gone the moment that session ends. As Richmond puts it, memory engineering treats long-term memory as first-class infrastructure.

There are several forms of agent memory - Richmond walks us through the four core ones, borrowed from human cognition:

🔹𝗪𝗼𝗿𝗸𝗶𝗻𝗴 𝗺𝗲𝗺𝗼𝗿𝘆: what the agent is holding right now, the active session and context window.
🔹𝗘𝗽𝗶𝘀𝗼𝗱𝗶𝗰 𝗺𝗲𝗺𝗼𝗿𝘆: what happened before. The past interactions and events it can recall.
🔹𝗦𝗲𝗺𝗮𝗻𝘁𝗶𝗰 𝗺𝗲𝗺𝗼𝗿𝘆: facts and knowledge. What it knows about you, your data, the world.
🔹𝗣𝗿𝗼𝗰𝗲𝗱𝘂𝗿𝗮𝗹 𝗺𝗲𝗺𝗼𝗿𝘆: learned skills and rules. How it actually gets things done.

Working memory is short-term. Episodic, semantic, and procedural are the long-term forms, and they’re what turn a one-shot demo into an agent that improves from experience.

𝗧𝘄𝗼 𝗽𝗹𝗮𝗰𝗲𝘀 𝘁𝗼 𝘂𝗽𝘀𝗸𝗶𝗹𝗹:
1️⃣ Free short DeepLearning AI Agent-Memory course
2️⃣ Oracle AI Developer Hub for hands-on agentic AI labs (notebooks, agentic RAG, reasoning architectures, full reference apps)

Comment ✨developer✨ & I’ll DM you the links

07/13/2026

Let’s see if I can cover the ML pipeline in 60 seconds ⏰😅

Machine learning isn’t just training a model. A production ML lifecycle typically looks like this:

1️⃣ Define the problem & objective
2️⃣ Collect and (if needed) label data
3️⃣ Split into train / validation / test sets
4️⃣ Data preprocessing & feature engineering
5️⃣ Train the model (forward pass + backpropagation in deep learning)
6️⃣ Evaluate on held-out data to measure generalization
7️⃣ Hyperparameter tuning (learning rate, architecture, etc.)
8️⃣ Final testing before release
9️⃣ Deploy (batch inference or real-time serving behind an API)
🔟 Monitor for data drift, concept drift, latency, cost, and reliability
1️⃣1️⃣ Retrain when performance degrades

Training updates weights.
Evaluation measures performance.
Deployment serves predictions.
Monitoring keeps the system healthy.

It’s not linear. It’s a loop.

And once you move beyond a single experiment, that loop becomes a systems problem.

At scale, the challenge isn’t just modeling ... it’s building reliable, scalable infrastructure that supports the entire lifecycle.

Curious if this type of content is helpful! Lmk in the comments & as always Happy Building! 🤍

07/10/2026

Jolene cover 🎶 experimenting with coding my instrumentals with strudel (it’s a free live coding environment!)

Been wanting to play around more with strudel for a while.
LMK if you try it out! 😊

07/08/2026

I’ve been wearing Meta glasses in general almost every day for over a year, & love the new design options + improved tech specs! So let’s unbox them… but first, here’s personally WHY I use them daily👇

1️⃣ Safety 🚦 I walk everywhere in NYC, and I love that I can take a call, listen to a podcast, or play my music and still hear everything around me. I’m crossing streets aware, not tuned out.

2️⃣ Being present with my family 💙 Honestly half the time I don’t even have a free hand… I’m pushing a stroller, holding bags of kids stuff, and my phone is in a totally different room anyway. But it helps me catch photos and moments I would probably miss and actually look into their eyes while I do it, instead of watching the moment through a screen.

3️⃣ Even at home 🏡 sometimes I just want to listen to something, but I don’t want to block everyone out with AirPods in. I still want to hear my kids, my husband, everything going on around me. And I went with the transition lenses so they’re good inside and outside

I got a first look the day they dropped and ordered them as soon as I tried them on. They’re light, and there’s even a tiny gem 💎

Here’s the more detailed specs:

These are officially the Meta Glasses by Kylie Jenner (the Starfire edition), and they come in black or tortoiseshell 🖤🐆

🔹 Now they’re actually adjustable to fit your face … a 3-way adjustable nose pad, bendable temple tips, and hinges that flex wider
🔹 The case folds super thin so it fits right into my smaller bags, and it even has a mirror inside
🔹 The metal nose bridge wipes clean so easily (makeup girlies, you get it 😅)
🔹 A dedicated button to pull up Meta AI or grab a photo hands-free
🔹 Powered by Meta’s newest AI, Muse Spark 🙌 🔹 A 6-mic array that picks up your voice really well
🔹 Prescription lens support
🔹 12MP camera with 3K video
🔹 Open-ear speakers
🔹 8-hour battery, plus up to 40 hours from the case

(And across the whole new line there are 26 style + color combos if you are looking for different style)

Address

3355 S Las Vegas Blvd
Las Vegas, NV
89109

Website

Alerts

Be the first to know and let us send you an email when Linda Vivah posts news and promotions. Your email address will not be used for any other purpose, and you can unsubscribe at any time.

Contact The Business

Send a message to Linda Vivah:

Shortcuts

Share

Category