Linda Vivah

Linda Vivah Building in the Cloud
Sometimes “Sing As A Service”
Founder
Tech, Career, AI

DASH 2026 was a blast! ✨ I first integrated Datadog back in my first DevOps role years ago, so seeing how far the platfo...
06/11/2026

DASH 2026 was a blast! ✨ I first integrated Datadog back in my first DevOps role years ago, so seeing how far the platform has come, especially everything on the AI side, was incredible!

💜 Observability in the age of AI is so crucial: Traditional observability meant watching software you wrote and understood, where you could trace a request and know exactly what happened. AI agents flip that. They make their own decisions and act on their own, so control around the model matters more than ever: being able to see what an agent did, why it did it, and gate its actions before they ship.

💜 It was also a full circle moment. I ran into a bunch of people I used to work with back in my media tech days, which made the whole thing feel even more special & event attended a session form my former manager in DevOps & a mentor in my career who absolutely crushed it & shared so much wisdom on observability in sports media tech

💜 It was also great meeting friends from IG in person! That is always the best part of these conferences. Marina Petzel and I filmed a bunch of content together, and we cannot wait to share it. Follow her for all things agent observability!

💜 I also had the incredible opportunity to interview Alexis Lê-Quôc, co-founder and CTO of Datadog, and I cannot wait to share that interview with you!

💜 The thing that always stands out about Datadog though is the people. I consistently find they have some of the nicest folks in the industry. Shout out to Madi Moore with The Coolest Nails Ever 💅

Machine learning isn’t just training a model. A production ML lifecycle typically looks like this:1️⃣ Define the problem...
06/10/2026

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! 🤍

Holding the new NVIDIA RTX Spark superchip which can run models with up to 120 billion parameters locally! 🤯I had the op...
06/06/2026

Holding the new NVIDIA RTX Spark superchip which can run models with up to 120 billion parameters locally! 🤯

I had the opportunity to visit the Surface lab during Microsoft Build, where this chip is powering the two new Surface devices Microsoft just announced this week: the Surface Laptop Ultra and the Surface RTX Spark Dev Box.

The specs 🤓
It pairs a 20-core Grace CPU with a Blackwell GPU, up to 128GB of unified memory, and up to 1 petaflop of FP4 AI performance. Which means the devices it powers can fine-tune models and run AI agents locally, without sending anything to the cloud.. wild!

I also didn’t expect the superchip to be this thin! Video on this visit coming soon. Thank you so much , & for the opportunity to visit the lab & get a first look!

05/21/2026

Claude Skills: 90 second mini crash course.. let’s go!✨

Skills teach Claude to do something the same way every time. They can be added in 4 different ways, live in 3 different places and the most fascinating part is how they load to avoid context bloat.

Here’s the TLDR:

4 ways to get skills 🔎
1️⃣ Built-in
2️⃣ Community (GitHub)
3️⃣ Custom (your own SKILL.md)
4️⃣ Plugins

3 places they live 🏡
1️⃣ Claude.ai
2️⃣ Claude Code
3️⃣ API

🧠 How they load?
Progressive disclosure, in 3 layers:

1️⃣ At startup, Claude only peeks at each skill’s name + description (~60 tokens). Just enough to know it exists.
2️⃣ When a task actually matches, the full SKILL.md loads.
3️⃣ Any extra reference files inside the skill? Only load if that specific task needs them.

7 core Claude Code features mini crash course….let’s go! ✨To unlock Claude Code’s full potential you need to know what’s...
05/19/2026

7 core Claude Code features mini crash course….let’s go! ✨

To unlock Claude Code’s full potential you need to know what’s available to you. Here’s the TLDR:
🔹CLAUDE.md → a markdown file Claude loads as context
each session
🔹 Skills → packaged instructions Claude loads when the task
matches
🔹 MCP → standard protocol for connecting Claude to external
tools and data
🔹 Hooks → commands that run automatically at points in
Claude’s lifecycle
🔹 Subagents → isolated Claude instances with their own
context and tool access
🔹 Plugins → skills, hooks, and MCP configs bundled into one
installable unit
🔹 Agent Teams → parallel Claude sessions with a lead,
messaging each other directly

Thank you for the color analysis prompt for ‘s new image 2.0 model - been wanting to do a color analysis & I tend to d...
04/28/2026

Thank you for the color analysis prompt for ‘s new image 2.0 model - been wanting to do a color analysis & I tend to default to black (true New Yorker 😂)

Been playing around with the new 2.0 image model & it’s incredible especially for images that combine text 🔥

✨Check out ‘s IG for the color analysis prompt

04/28/2026

Claude Chat vs Claude Code vs Claude Cowork - when do you use what? I got you👇

It comes down to two things:
1️⃣ How much access Claude has?
2️⃣ How much YOU want to stay in the loop vs delegate?

🍪 Quick preface though…. no one eats an Oreo the same way these 3 have overlap by design especially Claude Code and Cowork. Best way is to try it out

💬 Chat is what most people start with. You’re driving every step. Web search, connectors, MCP, the whole thing. It’s great for thinking together. And if you want to share a workspace with your team, Projects in Chat is still the only place where that’s actually shareable.

💻Claude Code is the most powerful, and honestly the most flexible. It lives wherever you do, terminal, IDE, the desktop app, the web, even your phone. Channels lets you text it from Telegram or Discord, which is wild. Full filesystem, full codebase, edits across files, runs tests, opens PRs. You can spin up agent teams in parallel. And since I filmed this, Routines also shipped, which means Code can now run on a schedule, an API call, or a GitHub event in Anthropic’s cloud, even when your laptop is closed. If the output is code, this is your surface.

💼 Cowork is on the desktop app. Same engine as Code, no terminal, no setup. You give it a goal and it works through your local files and apps and comes back with a finished doc, report, or spreadsheet. Dispatch feature lets you message it from your phone but the work still runs on your desktop.

✨ The coolest thing about AI is personalization & there are constant updates but there is no right or wrong. It takes time to find a routine and i would first approach it as experimentation with no pressure of choosing 😊

[AI engineer, anthropic, coding]

04/25/2026

💡For simplicity today we use “AI” as a catch-all but here’s a quick overview so you know the technical difference:

All GenAI is part of AI, but not all AI is GenAI

🤖 AI
Technology that enables machines to mimic human intelligence and perform tasks like reasoning, learning, and problem-solving

📊 Machine Learning
A branch of AI that focuses on building systems that learn from data and improve over time without being explicitly programmed.

🧠 Deep Learning
A subset of ML that uses neural networks with many layers to process complex data like images, audio, and text.

🎨 Generative AI
A type of AI that can create new content — like text, images, audio, and code …based on patterns it learned from training data.

04/23/2026

7 core Claude Code features explained in 90 seconds.. let’s go! ⚡

To unlock Claude Code’s full potential you need to know what’s available to you. Here’s the TLDR:
🔷 CLAUDE.md → a markdown file Claude loads as context each session
🔷 Skills → packaged instructions Claude loads when the task matches
🔷 MCP → standard protocol for connecting Claude to external tools and data
🔷 Hooks → commands that run automatically at points in Claude’s lifecycle
🔷 Subagents → isolated Claude instances with their own context and tool access
🔷 Plugins → skills, hooks, and MCP configs bundled into one installable unit
🔷 Agent Teams → parallel Claude sessions with a lead, messaging each other directly

04/22/2026

It’s their world… I’m just living in it 🤣❤️🎬

[mom of 3, women in tech, working mom, mompreneur, entrepreneur, AI engineer]

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