Decompute

Excited for today's AI Users Conference- Developer day! I will be joining Hina Chauhan Dixit, David Kanter , and Subha Tatavarti on a panel to talk about the future of AI on-the-edge! If you are attending the conference HMU 🤙 If you would like to sign up for the event, you can here: https://lnkd.in/dtZ4Tz-V

We are hiring for Head of Engineering - AI at Decompute. We’re building the future of on-device AI, where intelligence runs locally, privately, and efficiently. We're looking for a seasoned engineering leader to drive execution, scale AI infrastructure, and shape the next generation of intelligent systems. #hiring #AI

Excited to speak at the upcoming AI User Conference in San Francisco! I’ll be joining an incredible panel to discuss “Decentralized Intelligence: The Future of AI on Your Device” - exploring what it truly means to bring AI closer to the edge, into the hands of users, with speed, privacy, and control. Panelists include: • David Kanter, Founder of MLCommons • Aishwarya Srinivasan, AI at Fireworks AI • Subha Tatavarti, Advisor and Former CTO, Wipro It’s wonderful to share the stage with our early angels and supporters and talk about Decompute, where we’re building the infrastructure for on-device AI: training and inference that’s fast, private, and local. Join us: April 17th, 2025 · 4:30 PM · Mason Fort, SF #AI #EdgeAI #Decompute #OnDeviceAI #TechConference #DecentralizedAI #MLX #StartupLife #AIUserConference

Breaking Biases: A Female Founder’s Journey – Vol. 2 Let’s start with the data, because I’ve seen both sides of this game. Between 2014 and 2024, companies founded only by women received just 6.8% of first-time funding. In the same timeframe, male-only founding teams received 76.2%. And 2024? All-women teams raised just 2.2% of total venture capital. Yes, 2.2%. Now let’s talk about the lived experience. Being a female founder raising in deep tech isn’t just difficult, it can feel like a system stacked against you. If you’re struggling to get that “yes” while watching others with less traction or polish fly through, you are not alone. Here’s what’s happening beneath the surface: The Hard Stats: It’s Not You, It’s the System • Women-led startups receive <2% of U.S. VC funding. • In deep tech & infrastructure AI, women founders are even rarer. • We’re often told we’re “too early,” while male peers are called “ambitious.” • We’re asked for more traction, more validation, more perfection, while others get funded on slides and vibes. So if you’re being overlooked, hear this: It’s not because you’re not good enough. It’s because the system wasn’t built for you. But here’s the truth: rigged games can be hacked. I’ll never forget one VC telling me he passed on a brilliant female founder because she said, “We’re an underdog, but we’re here to win.” He didn’t like her using the word “Underdog”. That same VC? Threw money at Vibe Coders with zero originality. I’ve learned to spot technical smoke and mirrors. And I’ve learned this, too: perception is rarely about merit. Mindset Shift: Reclaiming Power When You Feel Forsaken • You’re not here to ask for permission. You’re building what the world needs, even if it doesn’t recognize it yet. • Don’t internalize rejection. If your own family couldn’t see your worth, how could strangers? Let that pain be your power, not your measure of value. Tactical Wins: Build Credibility Without a Check • Narrative > Capital You don’t need $5M in the bank to own your story. Say it boldly, on LinkedIn, on X, and in your pitch deck. • Flex Your Proof Got 100 users? A working product? A demo that beats Big Guys? Use that. Own that. • Create Your Own Validation Write. Speak. Publish. Teach. Build credibility through thought leadership, not gatekeeper approval. • Find Micro-Yeses Forget Sand Hill for a minute. Talk to customers. Builders. Students. If they say it’s amazing, that’s the signal. VCs follow traction, not feelings. If this hits home, message me. If you’re building and feel unseen, I’ve been there too. Let’s hack the system. Let’s win, on our terms. (Also, BlackBird macOS was downloaded 500+ times in the last 3 days, our website visits grew to 3000+ sessions) Breaking bias, one build at a time. 😎 #FemaleFounders #OnDeviceAI #StartupLife #DeepTech #VentureCapital #BiasInTech #WomenInAI

We are hiring! Looking for Head of Engineering! Please send resume to hi@decompute.run.

🚀 Revolutionizing AI Efficiency: Introducing Aurora 🌟 The future of AI is here, and it’s faster, smarter, and more efficient than ever. Meet Aurora, a groundbreaking optimization for Transformer architectures that’s redefining what’s possible in machine learning. 💡 Why Aurora Matters: Transformers are the backbone of modern AI, powering everything from chatbots to autonomous systems. But their reliance on Softmax has always been a bottleneck—slowing down computations and consuming vast resources, especially with long sequences. Aurora changes the game: ✅ Near-linear complexity for self-attention mechanisms ✅ Lightning-fast training & inference for long-context models like GPT and BERT ✅ Deployable on edge devices for real-time applications Whether you’re building NLP models, recommendation systems, or autonomous solutions, Aurora unlocks new levels of scalability and performance. 📈 Proven Results: Tests on models like DeepSeek Coder 1.3B and Llama 3.2 1B show dramatic runtime improvements—especially with growing context lengths. Imagine faster insights, reduced costs, and cutting-edge innovation at your fingertips. 🌍 Impact Across Industries: From NLP to computer vision to autonomous vehicles, Aurora is paving the way for AI to thrive in real-time environments while lowering resource demands. It’s time to embrace the future of machine learning efficiency. Are you ready to transform your AI workflows? 👉 Learn more about Aurora here: https://lnkd.in/gUBttqbP Let’s build smarter AI together! 💻✨ #AI #MachineLearning #Innovation #Transformers #Aurora #Efficiency #TechLeadership

The future of AI won’t live in the cloud. It’ll live on your laptop. Introducing BlackBird by Decompute: A new way to train AI models & agents — no datasets, no config, no cloud. Download for macOS here: https://lnkd.in/gepJy3xC Just drop in your unstructured data — PDFs, notes, folders, anything. BlackBird learns without needing labeled datasets. Zero setup. Full control. Total privacy. Your laptop becomes a personal AI lab. Launch post: https://lnkd.in/gCt8Yf9k Waitlist for Windows & iOS: https://lnkd.in/gbJ7FGuk

The cloud is not the future of AI. Today, we’re launching BlackBird 1.0 ( for MacOS, Windows shortly) - the most powerful way to build smart, autonomous AI agents that live entirely on your device. No cloud. No streaming your data across the internet. Just pure, private intelligence — local, efficient, and under your control. In a world where AI costs are exploding and privacy is optional, BlackBird delivers something radically different: • On-device AI that runs on laptops, tablets, and phones • No GPU farms, no cloud fees • Fine-tuned agents with the lowest memory footprint in the world • Full sovereignty over your models and data You don’t need a data center to have powerful AI. You just need the right platform. This is how we cut cloud costs, unlock edge autonomy, and make AI yours again. Welcome to BlackBird. Privacy first. Edge-native. Built for the real world. Launch post: https://lnkd.in/gYhE-kWJ #Decompute #BlackBird #EdgeAI #OnDeviceAI #AgenticAI #AIInfrastructure #LocalLLMs #MixtureOfExperts #PersonalAI #CloudlessAI #AITransformation #TechLaunch #PrivacyByDesign

At Decompute, we believe the future of AI is decentralized. LaserTune brings us one step closer by enabling efficient, high-quality fine-tuning without relying on centralized infrastructure. Built for: - Memory-constrained edge devices - Private, on-prem compute environments - Speed, scalability, and security by design Whether you’re tuning models on a laptop, a workstation, or an air-gapped server, LaserTune adapts to your local context—efficiently and privately. Read the full blog post here: https://lnkd.in/gMg4ViqH Let us know what you’re building—we’d love to collaborate for your fine-tuning needs and help you save 50% of the compute costs. #AI #EdgeAI #FineTuning #Decompute #DecentralizedAI #LaserTune #OnDeviceAI #PrivateAI

Thank you Dennis Salguero for sharing the inspiring journey of our CEO Hina Dixit. You can check out the full episode here: https://lnkd.in/gBzkktt4