Emad Hasan

If hardware is going to compress from multi-year cycles to adaptive, intelligence-driven programs, it won’t just require builders. It will require platform thinkers. Excited to welcome George Thangadurai to Wizerr’s Advisory Board. George is a Growth-Stage Partner at Avataar Venture Partners and one of our earliest backers. With 30+ years across enterprise software, AI/ML, cloud, and IoT, including leadership at Intel Corporation, Intel Capital and multiple IPO and M&A journeys, he understands what it takes to scale global infrastructure. At Wizerr AI, we are building the AI-native inference infrastructure for hardware systems, where engineering, sourcing, and supply decisions converge at the program level. “As an investor, I look for platform-scale opportunities with durable moats. Wizerr has the ingredients to build the intelligence backbone for hardware programs globally.” Grateful to build this with leaders who see where hardware is heading.

74

9 Comments

Product leaders Aatish Nayak from Harvey and Sachi Shah from Sierra joined us for Builders Ep 2 to talk about what it really takes to scale AI application companies. We cover: - Starting with “context complete” workflows - Evolving from Q&A to multi-user agentic systems - Merits and limitations of the FDE model - Building trust with transparency and UX - When to trust when “vibes are good.” Watch: https://lnkd.in/gjMUbWPr

60

4 Comments

Most of the AI tooling conversation right now centers on code generation. Cognition, Cursor, Claude Code....these tools are genuinely useful, but engineers are drowning in production issues. At our latest Greylock Partners Change Agents event, Mayank Agarwal and Rushin Shah from Resolve AI walked through a reality most engineering teams live through: with AI coding tools "you can crank out code much faster, but teams aren't really shipping to production that much faster." Some takeaways from our conversation: 1/ The real problem isn't code generation. Mayank saw this at Splunk where engineers spent only 10-20% of their time writing new features. The rest was fighting incidents, navigating tribal knowledge across hundreds of microservices, and operating tools that don't talk to each other. 2/ Production systems are different from code. Your codebase is static. Production is emergent behavior from infrastructure, deployments, traffic patterns, to tribal knowledge buried in Slack. As Mayank stated, the issue is "a coding agent cannot predict these things without the deep understanding of what's actually happening in production." 3/ The scale required is massive. Resolve is now consuming a trillion tokens per quarter—1% of all Azure usage. That's running Stack Overflow from scratch twice per week. "Vibe debugging" is the missing piece, Rushin's term for getting clarity on production systems the way vibe coding accelerates writing code. Check out the full episode in the comments

104

28 Comments

From Measurement to Orchestration, with Pranav Piyush from Paramark.com In this week's Intelligent Marketer ep, Rishabh and I dug into the evolution of marketing measurement in a world where direct, click-based attribution has broken. The smartest marketing teams are building systems that triangulate incrementality testing + MMM results. Pranav discusses how getting this "central brain" right is a critical foundation for AI-driven orchestration & execution of an end-to-end marketing roadmap. Topics covered: - Policies, not dashboards: how to decode decisions as machine-readable rules that turn insights into actions - Confidence-weighted budgets: how to use error bars in lift/MMM results to drive budget reallocation - Latency-aware orchestration: how to incorporate daily platform signals vs weekly incrementality results vs quarterly MMM output - Creative as a portfolio: how to weight hit rate & EV alongside typical ROAS metrics - Constraint-aware bidding: how to bake margin, inventory, compliance, and brand safety costs into the decision function

40

4 Comments

Agentic AI - The Next Engineering Frontier The shift from reactive AI to proactive agents is accelerating faster than most anticipated. We're moving beyond single-task models to autonomous systems that can plan, execute, and adapt—fundamentally changing how we architect intelligent systems. Key trends shaping 2025: - Multi-agent orchestration replacing monolithic deployments - Tool-calling capabilities becoming standard across LLMs - Reasoning chains enabling complex problem decomposition - Human-AI collaboration patterns emerging as the sweet spot The roadmap ahead involves solving coordination complexity, establishing safety guardrails, and building robust evaluation frameworks. Benchmarks like GAIA remind us how challenging true general intelligence remains—even for seasoned engineers. As we scale these systems, the focus shifts from "can we build it?" to "how do we govern it responsibly?" The engineering challenges are immense, but so is the potential impact. Recently completed Hugging Face excellent agentic AI course—their practical approach to complex concepts is impressive. The hands-on framework really crystallizes the architectural patterns we're seeing in production #AgenticAI #Engineering #MachineLearning #TechLeadership #AIStrategy

64

8 Comments

Wondering what Forward Deployed Engineers actually are and when it makes sense to use them? I recently had a great conversation with CJ Gustafson on FDEs, and how we use them at Invisible Technologies. We spoke specifically about FDEs as a GTM choice, how the economics change when engineering is pulled forward into the sales motion, pricing models, metrics to measure health of business and lots of other things. Check out more on the latest Mostly Media newsletter. If you're not subscribed, highly recommend subscribing. Link in comments.

40

5 Comments

Ramblr.ai partners with Nakamir to bring #PhysicalAI to the US Navy ! We’re excited to share that Ramblr.ai will be partnering with Nakamir on a newly awarded #SBIR project with the US Navy, focused on bringing next‑generation #AI‑powered work guidance into Navy shipyards. As part of the effort, Ramblr.ai will deliver advanced computer vision for real‑world video understanding, enabling shipyard teams to capture expertise, improve task execution, and accelerate maintenance workflows through intuitive, #AI‑driven guidance. Shipyards everywhere face increasing demand, complex maintenance requirements, and a persistent shortage of skilled labor. We believe #AI built on real-world video and #AR data can be a major force multiplier — and this program is a big step toward that future. Christoph Leuze - thanks for the trust - let´s ship it :-) https://ramblr.ai #DefenseInnovation #AI #Shipbuilding #AugmentedReality #SBIR #WorkforceDevelopment #ComputerVision #SmartGlasses

156

13 Comments

AI scaling doesn’t stall on compute. It stalls on data. Epoch AI’s recent analysis makes that clear. We can build bigger models, fabricate more chips, and wire up more power. But the supply of authentic, high-quality human data is finite—and exhaustion is coming fast. Synthetic data helps, but at scale it risks “model collapse.” For autonomous vehicles, this is the long pole in the tent. Models can’t just learn from themselves. They need to see the messy, diverse, unpredictable patterns of the real world—pedestrians stepping into traffic, trucks cutting across lanes, work zones appearing overnight. This is where Nexar’s role in the AV ecosystem is critical. Our network of hundreds of thousands of dash cams streams back the ground-truth data that AV models actually need: fresh, varied, real-world scenarios, at scale. The industry can’t compute its way past the data problem. Scaling AVs will depend on who can deliver authentic real-world signals—and how responsibly we build them into the ecosystem. https://lnkd.in/eYCZreZE

19

1 Comment

Barbarians at the Gate (https://lnkd.in/gkEBX79s) is a very interesting new paper from some great folks at Berkeley, with some exciting results showing the potential for AI in systems research. But I think the authors aren't quite asking the hardest problem about where this takes systems as a field. I wrote a new blog post about it: https://lnkd.in/gxS3s4v4 The core point of the paper is that AI can be effective at automatically driving the 'number go up' part of systems research. Given an evaluator, AI can automate progress on implementation. I believe this is 100% correct. The paper also talks about how more focus will have to go into developing evaluators. Again, I 100% agree, but think they're understating their case. Building evaluators (basically tests) is, I believe, going to become the focus of systems and software over the next decade. Which leaves researchers with the hard problem of choosing which problems to go after. That's never been easy in academic systems, and is only going to become more of a challenge. But the best systems research is driven by choosing problems well! This vision for AI could lead to lots more of the least interesting systems research. The "number go up" kind. Which could make it harder than ever to identify the truly innovative and visionary work. We're already seeing that elsewhere in science. So, mostly, I think the paper is right, but not yet asking the biggest question: what should systems research be?

276

12 Comments

Reasoning models can generate up to 20 times more tokens than non-reasoning models, although recent non-reasoning models also tend to produce increasingly large outputs. Despite many brilliant ideas emerging from the community, the decoding phase remains fundamentally memory-bound. This makes compression-oriented AI chips essential for building sustainable and affordable AI services—particularly to support the growing demands of agentic AI. At NAVER Cloud, this is why our team has been actively advancing both novel AI compression and optimization techniques (published at ICLR, ICML, NeurIPS, and others), as well as designing compression-oriented AI chips (published at MICRO, HPCA, SC, ISCA, and more). Source: https://lnkd.in/gd-iSaKF

109

3 Comments

Forward-deployed engineers (FDEs) have become one of the most in-demand roles for AI companies. Everyone's racing to adopt the latest AI models and applications, but here's what we're seeing: the biggest differentiator for AI companies isn't what model(s) they're using, it's how well they're implementing them to solve their clients' actual use cases. SaaS was about digitizing business processes. This new era of software is about using agentic AI to rethink and automate them. Because AI applications are systems of action (shoutout to Unify for coining this), they often need FDEs to translate a customer's messy business reality, not just their documented SOPs, into applications that perform reliably in production. If AI models are the gold, then forward-deployed engineers are the miners. Our partner Gordon Ritter and I wrote more about why FDEs are becoming one of the most strategic roles in enterprise AI and what second-order startup opportunities this creates. Would love to hear more thoughts on how you're seeing this trend play out! https://lnkd.in/g-gaBZeb

111

6 Comments