Faiz Abdul

Having spent the last two days at QCon, I had several chats with engineers who wanted to know what 'AI Engineering' as a role meant. Pete Hamilton's written about why we've created the role which is where I pointed most people. The tl;dr is you do different work than Product Engineering, and we're betting on the specialisation only becoming more relevant in future. Explaining this to people, their next question was "that makes sense, but why would I switch? what would this mean for me, in terms of my career?" When I switched role to SRE, Norberto (my manager at the time) gave me a list of pros-and-cons around what it would mean for my career. He emphasised parts of the role that would develop certain skills, and was honest about what I'd lose moving away from normal engineering. I've tried writing-up a similar list for anyone considering the switch to AI Engineering, hoping to return the favour! It covers: - Why AI engineering is one of the most powerful career accelerants available right now - What you'll learn from battling AI's technical challenges - That you'll build less visible product than you might in other roles and what that means Link to the post is in 🧵

38

3 Comments

Transitioning from engineering to product during the rise of LLMs has been a pretty interesting experience - I wrote some reflections and predictions for what's coming! ---- “𝐈𝐭'𝐬 𝐨𝐧𝐥𝐲 𝐰𝐡𝐞𝐧 𝐭𝐡𝐞 𝐭𝐢𝐝𝐞 𝐠𝐨𝐞𝐬 𝐨𝐮𝐭 𝐭𝐡𝐚𝐭 𝐲𝐨𝐮 𝐝𝐢𝐬𝐜𝐨𝐯𝐞𝐫 𝐰𝐡𝐨'𝐬 𝐛𝐞𝐞𝐧 𝐬𝐰𝐢𝐦𝐦𝐢𝐧𝐠 𝐧𝐚𝐤𝐞𝐝” - 𝐖𝐚𝐫𝐫𝐞𝐧 𝐁𝐮𝐟𝐟𝐞𝐭𝐭. As the tide pulls out on much of coding and knowledge work, what actually matters when building products? If 2026 pans out the way we're all predicting, I suspect a lot of engineers (and product managers!) will find out they've been swimming naked. Last year, I just so happened to be one of them.

68

4 Comments

Our team published a deep dive on how we built our in-house AI data agent to work at real data-platform scale. The agent reasons over 600+ PB across ~70k datasets, using natural language to help engineers, product teams, and researchers get trustworthy answers grounded in correct metadata, ownership, and access controls. The real challenge wasn’t just LLMs, it was building tight integrations with our data platform: table-level knowledge, catalogs, governance, and execution layers, so answers are accurate, fresh, and auditable. Under the hood: Codex-powered reasoning on top of structured metadata and platform context. Learn more on our Eng Blog: https://lnkd.in/gvCCbhrk If you're interested in building bleeding-edge AI tools for data analysis, please apply today: https://lnkd.in/g7DUrUNz

270

17 Comments

Week 1 as a Forward Deployed Engineer(#FDE) Here’s how I see the role👇 No long blog this time, just quick notes: 🔹 Customers don’t buy an “AI Agent product”. They buy a better solution to make their end users happy, make their team productive, stronger branding and lower cost 🖌️ Think of a pencil company. Selling the pencil (product) isn’t enough. The customer wants a portrait hanging in their home (solution). That takes a team who can understand the vision, sketch it out, and make it real. In AI, that’s the FDE 🎨 Regular SaaS or OnPrem software are selling the pencil, or even fancy painting machines. But in the AI era, with LLMs, determinism is an illusion. Delivering value takes engineering, science, art — and sometimes a little luck 👩‍🎨 You can invest in the most powerful “Agent Platform,” but customers still need help using it to deliver a real AI experience. That’s where FDEs come in. 🔵 Blue pill: 90% platform, 10% FDE. 🔴 Red pill: 50% platform, 50% FDE. Which one would you take? For customers, the split doesn’t matter — what matters is whether the solution actually works That's my takeaway for the first week at Cresta, a leading AI Agent company. After the usual IT/HR onboarding,  I was already on-site with a major customer by Day 2. I learned a ton from other FDE, PM, Sales and the customer, and it feels like I’m just getting started Curious about the FDE role in the AI era? Let’s chat — and yes, we’re hiring

54

4 Comments

In this week’s AI infra chat, we sit down with Seiji Eicher, Distributed LLM Inference Engineer at Anyscale to unpack how large MoE models are actually served in practice, what it looks like to work deep in the LLM inference stack & more 🕐 Timestamps: 𝟬𝟬:𝟬𝟬 Why MoE comes up in modern AI infra 𝟬𝟬:𝟮𝟯 What Mixture of Experts actually changes inside a model 𝟬𝟭:𝟮𝟰 Why expert parallelism exists (and what becomes inefficient without it) 𝟬𝟮:𝟱𝟳 When MoE helps in practice 𝟬𝟰:𝟮𝟯 How Ray Serve fits into real inference stacks 𝟬𝟳:𝟬𝟳 What a Distributed LLM Inference Engineer works on day to day 𝟭𝟭:𝟭𝟵 Where the space is headed and how to learn more If you want to go deeper, Seiji gave a Ray Summit talk that dives into production-grade serving of MoE models, including KV-cache optimization, parallelism strategies, and multi-node orchestration with Ray Serve and vLLM. Check out the talk linked in the comments 👇

60

3 Comments

✨ Introducing: Modal Notebooks ✨ Really excited to share what I’ve spent the last 8 months working on—Modal Notebooks. It’s the latest product built on our AI infrastructure, and it reimagines the classic notebook DevEx from the ground up. Modal Notebooks is a collaborative environment for high-performance interactive computing. We’re purpose-built for AI research and experimentation: ⚡ Fast container startup (<5s cold start) on arbitrary images 💥 Instant GPU access, up to 8× H100/B200 GPUs 👥 Real-time collaboration with multiple cursors, live presence, and shared Jupyter Widgets 🛠️ Modern developer features — AI completions, semantic highlighting, Pyright LSP, rich outputs 📦 Unified environment with Volumes, Secrets, and deployed Functions from your Modal Apps 💡 Smart economics — kernels auto-idle and resume, so you only pay for active compute Why did we build this? Jupyter notebooks are invaluable, but today’s cloud versions are slow to start, hard to share, and limited on GPUs. Modal Notebooks remove those points of friction so teams can focus on research, iteration, and building. In the time of our short beta, 5,000+ accounts ran 200,000+ code cells with instant startup and real-time collaboration. Early adopters like Suno, sync., and Ligo Biosciences have been using Modal Notebooks to accelerate their research workflows. Developers already love using Modal to deploy AI workloads, and with Notebooks, we’re making research and exploration a delight, too. Full announcement in the comments below 👇

633

43 Comments