Altara

How will AI reshape how frontier technology is built – all the way from scientific discovery to scaled manufacturing? I’m excited to join the Innovation Research Interchange for a webinar on July 16th to dive into this question. We’ll cover insights on legacy data, the complexities of real-world AI adoption, failure analysis, and the practical challenges of accelerating R&D without ultimately creating new bottlenecks elsewhere in the product development lifecycle. If you're interested in the intersection of AI, industrial innovation, and scientific discovery, I'd love to have you join us. Registration link below.

Last week I interviewed Mike KriegerMike Krieger. Here are 5 lessons on product and company building he shared: 1. Close the gap between model capability and product experience. Right now, AI products can feel too constrained. The goal should be to allow more freeform user workflows. Mike gave an example that sometimes it’s as simple as addressing a user's frustration over not being able to alphabetically sort project names. The model is powerful enough to do so much more, but the product experience limits what’s possible. 2. UI is just the tip of the iceberg. Mike described UI as a “lossy compression” of all the underlying design and product docs, discussions, and decisions. To ensure you aren't losing the core value, continuous validation with your customers is non-negotiable, regardless of what you're building. 3. Not every prototype needs to ship. There are a ton of cool ideas and prototypes internally at Anthropic that have not shipped. Just because it’s easier than ever to ship code today doesn’t mean every prototype should ship. 4. Keep the human voice in product decisions. At Anthropic, there's a strong emphasis on maintaining authenticity. In product docs, teammates explicitly share how AI was involved in the product decision making process (ex: "I wrote the first pass, Claude edited it"). 5. Never lose your "hands-on" edge. No matter how large the organization gets, staying close to the craft matters, especially in this AI native era. Earlier this year, Mike stepped down from Chief Product Officer to be a hands-on IC at Anthropic Labs. Even as the team grew exponentially during his time as the CTO of Instagram, Mike still made it a point to ship code.  Huge thank you to Mike for sharing his wisdom and to NeoNeo for hosting a great event. At AltaraAltara we’re building a product designed to bridge frontier technology with physical world impact. If you're a software or research engineer who wants deep ownership over these kinds of complex product challenges, we're actively hiring: altara(dot)co/careers

One of our users recently DMed me: “A few minutes using Altara got me more excited about LLM-based tech than hours of keynotes from OpenAI and Anthropic. It's really refreshing to actually experience real transformation as opposed to endless hype.” Moments like these are why Catherine Yeo and I started Altara – because we believe that scientists working on the hardest problems deserve exceptional software. Too many scientists and engineers are stuck using tools stuck decades in the past, and it’s one of the largest bottlenecks preventing frontier technology from being developed and brought to market. Messages like this are also a great reminder of the importance of domain-specific AI for domain-specific problems. We’ve built Altara from the ground-up to be optimized for the use cases encountered by our customers across semiconductors, advanced materials, and batteries. If you’re interested in seeing a demo, request on our website at altara(dot)co or DM me directly.

May was a historic month for Altara. Quick recap: 1) Officially launched out of stealth with our $7M seed round led by Greylock Partners. 2) Record-high usage from early customers. 3) Named to the Black Flag 100 deep tech frontier list. 4) I partnered with OpenAI on a Founders Day video feature highlighting how we scale scientific intelligence. 5) Eva Tuecke spoke on stage at SciFM 2026 on deploying autonomous AI agents into real-world industrial environments. We’re only halfway through June, and this month is moving even faster. We're hiring across engineering, design, and GTM to keep this momentum going: altara(dot)co/careers

Recently joined Altara! I'm hyped to work on the bridge between physical sciences and AI. The combination of exceptional talent, obsession, and rapid customer feedback loops has made for an exciting start. We're hiring exceptional people who thrive in high-agency, fast-moving teams. Roles are in-person in SF. Reach out if that sounds like you! 🚀

So excited to share that I'm just wrapping up my second week at Altara! ⚡ And honestly, I couldn't have asked for a better team to be building with. 🔬 Altara builds a single platform that brings scientific R&D data together for semiconductors, batteries, and advanced materials. Scientists, engineers, and their AI agents get one place to explore data that's usually scattered across spreadsheets, equipment logs, and lab systems and accelerate their most critical, complex workflows 🧪 ⚠️ We're hiring in SF! ⚠️ If you're excited by hard problems at the intersection of applied AI and real-world science, and want to build with a small but cracked team (Catherine Yeo, Eva Tuecke, and more!), my DMs are open.

I recently spoke at #SciFM 2026 about AI agents for science and four key considerations for deploying AI into real-world industrial environments. 1. Value of existing and legacy data:  At Altara, we work with companies sitting on decades of incredibly valuable technical data. That data is important context for decision making in critical scientific and engineering workflows, if only you can build systems that can reason across that long-tail, domain-specific knowledge. 2. Integration complexity:  Across semiconductors, batteries, and advanced materials, some of the hardest problems go beyond models – they’re constrained by infrastructure. Data lives across on-prem systems, disconnected databases, proprietary software, and countless spreadsheets. AI only becomes useful after you solve for the realities of integrating with real-world workflows and infrastructure. 3. Importance of last mile accuracy: Our customers use Altara to make incredibly high-stakes, physical-world decisions: which experiments to run, how to change manufacturing processes, and which materials meet quality standards. These stakes mean 95% accurate isn’t accurate enough, ultimately driving different product requirements. 4. Trust, transparency, and interpretability:  Scientists and engineers need interpretable and transparent systems even more than most users. That means clear provenance, transparent reasoning, cited data sources, and the ability to understand every calculation. These remain some of the biggest challenges – and opportunities – for bringing AI into real-world scientific and industrial workflows, and we’re excited to be going after them at Altara.

Excited to be speaking this week at SciFM 2026 in Chicago, a conference focused on AI agents for science. At Altara, we spend a lot of time thinking about what it takes to make AI systems useful in real scientific and industrial workflows – not just generating insights over curated datasets, but helping to drive meaningful decision-making in environments where experiments, simulations, instruments, and sensors are continuously producing new information. Much of our work sits at the intersection of agentic AI, multimodal integration, multi-step scientific workflows, and complex real-world infrastructure. I’m looking forward to diving deeper into these topics on May 29 as a panelist for “Closing the Loop on Real-Time Physical Systems: Feedback, Latency & Autonomous Decision-Making,” and learning from the broader conversations happening at the conference around the future of AI for science. If you’ll be at SciFM (link below) or in the Chicago area this week, let’s connect!

Altara’s AI aims to diagnose failures and help speed up R&D by unifying data siloed across spreadsheets and legacy systems.