Salil Deshpande

I’ve been working on the theory of why small, shippable factories are the perfect inversion to the centralized prefab factory model. It all collapsed into one line of math (credit to Gilles Retsin): Factory Efficiency = Usage × (Throughput × Product Value ) / ( CapEx + Deployment OpEx) The idea is that the closer you can drive Usage to 100 % while keeping the denominator tiny, the faster the flywheel spins.  A shippable microfactory tackles this by primarily focusing on a high Usage factor. Because the microfactory can be packed up afterwards and reused, its CapEx gets amortized over multiple projects, effectively raising its overall Usage across the year. If one project alone doesn’t fully occupy the factory, it can simply roll over to another job to maintain high Usage. Its CapEx is often an order of magnitude lower than a centralized factory (< $1M vs. $10-50M min), so the breakeven throughput is more achievable on a small pipeline of work. The real breakthrough here is portability. There is near-zero stranded capital and almost no idle time. To flesh this theory out, I wrote 3000 words on the topic for Brad Hargreaves and Thesis Driven. The article breaks down: - why “fixed factories, shipped goods” is being inverted to shipped factories, fixed goods - what a sub‑$1 M robotic cell does to CapEx per home versus a £45 M off‑site plant - early production/cost data on a shippable microfactory from Mollie Claypool, Gilles Retsin, Sam Baker and the Automated Architecture (AUAR) team - the hybrid reality of robots tackling the heavy structural components and human crews assembling faster and cheaper - technical constraints to pull this model off and future considerations on the optimal business model (own vs. rent) Major shoutout to Gilles Retsin for helping me pull this together. https://lnkd.in/eRTnbZPa

54

2 Comments

As AI workloads eat up global computing supply, DRAM prices are surging like never before—creating major challenges for teams trying to plan out their data center spend. One of my portfolio companies, Mext, is addressing the largest cost component in the datacenter: server memory (DRAM). From being on the board of Redis for the last twelve years, I’ve learned a lot about the various aspects of this problem and approaches to solving them. Mext found a breakthrough, using new AI techniques, for dramatically reducing the amount of server-DRAM required to run applications, all while maintaining performance. It intelligently manages the server’s memory, keeping hot pages (i.e., those in use by applications) in fast DRAM and offloading cold pages (i.e., those less used) to a much less expensive memory tier (e.g., NVMe Flash). Key to the approach is ensuring cold memory pages that are about to be accessed by an application are back in DRAM before the application needs them or notices that they were gone. This is done without modifications to the application or the OS – so it can run in the cloud or on-premise. Mext thus allows applications to either run using less DRAM or keep their DRAM footprint but do more with it. They're hosting a webinar along with Fred Weber (former CTO of AMD) on Jan 22nd at 10am pacific time. You can register here: https://lnkd.in/gXe9GCeE

153

3 Comments

GPUs have gotten faster, and models have gotten larger. But storage is still stuck in the past. Modern AI workloads need instant access to huge datasets. Instead, teams waste time on slow retrieval, idle GPUs, and costly pipelines. Archil is changing this. It delivers object-storage scalability at block-storage speeds. The platform allows for seamless, high-performance access to massive datasets. By eliminating cold starts and accelerating throughput 30X, Archil enables teams to fully utilize their infrastructure and move faster on training, analytics, and AI deployment. Archil’s founder Hunter Leath brings rare technical depth and operating experience from building Amazon EFS and optimizing Netflix’s cloud performance. This is a team building the foundational layer that AI workloads desperately need. I’m thrilled to lead Archil’s seed at Felicis with Nancy Wang and support their mission to reinvent cloud storage for the AI era. https://lnkd.in/gWueXhzz

65

2 Comments

Cerebras is now a generational AI company. But between 2018 and 2019, one tiny component nearly killed the company. Back then we were deep into building the first wafer-scale system. Everything depended on a single part most people will never notice. The interposer. It sits between the main board and the dinner-plate-sized silicon wafer. It has to carry enormous power and data loads while having to survive massive thermal cycling. Cold to hot. Off to fully lit. Over and over again. We had just built the largest chip in the world. There was no existing supply chain for it so we had to single source this part. That alone is terrifying. One vendor, no backup. If it failed, the system failed. If the system failed, the company could have failed. For almost three years, this thing was the boogeyman in every board meeting. Would it maintain electrical continuity? Would it warp under load? Would it survive burn-in? Would it work at scale? For long stretches, the honest answer was “we don’t know.” We were burning millions of dollars per month. Time was not our friend. A redesign could have pushed us far enough back that the market might have moved on. At one point, we pulled in a group that I first worked with at IDEO and that used to be called "Failure Analysis". That is not who you call when things are going well! We went back to first principles. Physics. Materials science. Coefficients of thermal expansion. Power dissipation per square millimeter. Analyzing what happens when you push this much power through a system at these temperatures. Again and again. Slowly, painfully, we got there. Finally the interposer worked. Every company pushing the frontier runs into problems like this. There are rarely quick, clever fixes. You just have to do real failure analysis, over and over, until the answer starts to reveal itself. The Cerebras team is itself built from the strongest, toughest material I’ve seen. They earned every inch of the other side of that problem.

335

26 Comments

Chatting with ZEDEDA CEO, Said Ouissal, and he has an elegant way of explaining the evolution of industrial Edge. Zededa provides OS for orchestrating Edge AI. 2010s were about IoT/Edge Connectivity. 2020s were about Edge computing/processing. 2030s will be all about Edge AI.

36

4 Comments

With RL techniques like GRPO, it’s become easy to fine-tune small models to outperform foundation models on vertical tasks. We’re open sourcing Osmosis-Apply-1.7B: a small model that merges code (similar to Cursor’s instant apply) better than foundation models. Links in comments! Using foundation models for high specificity, low complexity tasks like code merging can be overkill - instead, it’s better to use a specialized model instead for higher performance and latency. We fine-tuned the model on 100K real-world code commits where the model needs to merge a code edit snippet into an original code snippet. On the test dataset (subset of CommitPackFT), foundation model performance ranged between 0.77-0.93 reward score. In comparison, Osmosis-Apply-1.7B achieved a 0.98 reward score - while also being 3X-5X more affordable than the next cheaper model AND ~10X faster! Why this matters: the best AI agents use multiple models to ensure task specialization and reliability at scale - on top of cost and latency benefits as well. Links to the model and full write-up (with reward function, hyperparameters, etc.) are in the comments - reach out if you’re interested in using reinforcement learning!

239

18 Comments

"Build the event you wish existed" That's what we at Monadical did last week at #NYTW. We wanted an AI Engineers discussion for Engineers. There's always a lot of events out there for VCs to network, or for startups to learn about G2M, but very little on lessons learned from actual engineers in the field. So that's the event we hosted. Big thank yous to our awesome panel, Roy Pereira, Ben Cohen and Corey J. Gallon. Here's the key takeaways and the AI tools we're using. 🚀 All three panelists independently called AI Agents the most transformative LLM application they’ve used. They specifically called out Claude 3.5 Sonnet for its accuracy and reliability. 🪨 We dug into how LLMs are “jagged”, not general. They can be shockingly good at some tasks and completely fail at others. Everyone agreed: good evaluations are critical (and hard.) 🧪 Corey noted how public benchmarks and reality are two different things. Most public evals are saturated or gamed. ♊ Ben emphasized that AI projects are actually two projects: building the tool and building the evaluation process. 🧱 We explored how falling dev costs may impact startup defensibility and labor demand. Roy shared that founders are already shifting strategies in response. ⚒️ In a world of daily AI launches, the panel discussed how they decide what’s worth attention, and what’s just noise. They called out tools like Goose, Aider, Claude Code, and Monadical’s own Cubbi, which helps run agentic workflows safely in dev environments. (links in the comments). CTA: What would you want to hear in an AI Applied Engineering talk you attended?

33

5 Comments

Last year, we invested in Cognichip: what we expect to be a generational semiconductor company by making chip design as simple and accessible as vibe coding. They came out of stealth today with a vision for how we will design chips in the future. Today, building chips is nowhere nearly as simple as uttering these commands. In fact, this very task requires going through thousands of pages of documents and hundreds of hours of experienced engineer time. It gets worse with every tech node. Chip design is tedious, which is why chips cost a fortune to design, and take forever to bring to market, resulting in only a handful of (giant) chip companies. Faraj Aalaei, who personally built, grew, and took several chip companies public over the past 3 decades, founded Cognichip with Simon Sabato and Ehsan Kamalinejad to disrupt this status quo: https://lnkd.in/gAkM_k6Q He sounded the alarm a decade ago, and as a dear friend of Lux, he was kind enough to share his vision that became Cognichip: Cognichip has attracted Olympiad-medalist mathematicians and veteran chip designers to make chip design as accessible as coding apps. We expect Cognichip to enable a future where there are as many chip companies as there are software companies, rapidly accelerating compute. We are excited to embark on this journey with the fabulous team Faraj has assembled, along with our friends at Mayfield and FPV Ventures

121

11 Comments

We at Sound Ventures are incredibly excited to share today we led WitnessAI's $58M raise, with participation from Fin Capital, Qualcomm Ventures, Samsung Ventures, and Forgepoint, alongside existing investors Google Ventures and Ballistic Ventures… And it is just the beginning 🙌 The primary barrier to enterprise AI adoption isn’t tech debt; it’s tech doubt. WitnessAI has proven it can instill confidence in AI adoption, and has done so for multiple global leaders. We are thrilled to support the team as they use this capital to expand into new markets and solidify their position as the leading standard for safe, secure enterprise AI. In our diligence, we spoke with dozens of Fortune 500 security leaders and heard a clear theme: teams want to move fast with AI, but need protections that scale. WitnessAI provides that through a full, end-to-end security platform spanning AI applications, models, and agents, seamlessly integrating into a company’s existing infrastructure. As AI moves from experimentation to execution, WitnessAI enables organizations to deploy AI broadly and responsibly. Led by Rick Caccia, a world-class operator with over two decades of experience building and scaling security businesses, WitnessAI is on a path to define the category for enterprise AI security.

476

38 Comments

Generative AI’s greatest potential lies in the physical world, where the ability to simulate, design, and iterate faster can reshape trillion-dollar sectors. That’s why we’re thrilled to announce that N47 is leading Luminary Cloud’s $72M Series B. The Luminary team is pioneering the Physics AI Model Factory, underpinned by a GPU-native physics simulation platform. This technology is designed from the ground up to unleash the performance of modern GPUs and generate the high-value datasets that power next-generation AI. Now, engineers can run hundreds of design iterations overnight instead of waiting days. AI models trained on physics-grounded data can suggest new aerodynamic designs or optimize energy systems in ways previously impossible, and companies across industries can accelerate their innovation cycles while reducing cost. With co-founders Juan Alonso and Pete Schlampp, Luminary has the technical depth and operational experience to define the new era of Physics AI. We’re proud to partner with them as they create the future of simulation and redefine what’s possible in the built world. Learn more about Luminary Cloud’s vision here: https://lnkd.in/gbjeWrvy 

119

6 Comments

Inertia is the commercial fusion energy company, and today we're taking a big step forward with $450M in funding! With support from Bessemer Venture Partners GV (Google Ventures), Modern Capital, Threshold Ventures, Neo, Uncork Capital, Long Journey, and WndrCo we're taking proven physics from the lab to build the industrial foundation for gigawatt-scale fusion power plants. The road to get here involved $30 billion in government funding for inertial confinement fusion experiments, the groundbreaking demonstrations of fusion that works at the National Ignition Facility at Lawrence Livermore National Laboratory, the incredible ingenuity of my co-founders Andrea (Annie) Kritcher and Mike Dunne, and the team of DO-ers we’re building at Inertia. Now, with our funding, we're going to: 🔆 Build and scale the production of the world's most powerful laser  🔆 Build the world's first fusion fuel factory to mass produce fusion targets 🔆 Design and build the first gigawatt, utility-scale fusion energy power plant We're bringing star power to Earth. Read the founders' letter: https://inertia.com/news/ Read the press release: https://lnkd.in/gVEEYEpv And come build with us: https://inertia.com/jobs

1,176

116 Comments

We’re excited to announce that we have led Scala’s Seed round today as the company emerges from stealth.   We’ve known Rajeev Singh and Ardie Sameti for years. Across multiple companies and operating roles, Raj has demonstrated judgment and vision at scale. His ability to navigate complexity, build durable systems, and grow companies amid technological shifts. Ardie brings deep operator experience inside large-scale contact center environments, with firsthand insight into how fragmented tools obscure understanding and make decisive action harder than it should be. Scala grew out of lived experience inside the systems it is now reimagining, not from abstract theorizing.   Contact centers are becoming one of the earliest proving grounds for how humans and AI will operate together inside real organizations. As AI accelerates the speed and scale of work, complexity increases rather than disappears. Leaders have more data than ever, yet less shared understanding of what is actually happening across systems, teams, and automation. Against this backdrop it has never been more valuable to have an intelligence layer that helps people reason about their operations — and enable deliberate action.   Scala is building that layer and we are excited to join our friends at FUSE in backing them at day 1.   In an era defined by reasoning systems and rapidly compounding change, founders who combine domain insight with technological imagination matter more than ever. We’re proud to partner with Ardie, Raj, and the Scala team for the long journey ahead. https://lnkd.in/gfGpP8Ue

113

11 Comments