RT @DisruptAfrica: SA’s Cerebrium raises $8.5m funding to scale leading high-performance serverless AI platform

This is the first version and there’s a lot more coming so feedback is appreciated!.

What you can do:.- Run unit tests on your app.- Trigger one-off jobs (Compiling TensorRT engine, preprocessing etc).- Use your secrets and access persistent storage.- Get realtime logs.- Attach multiple CPUs/GPUs. It’s like your local dev loop — but in the cloud.

🚀 Launch Week Day 2: Introducing cerebrium run. One of the biggest pain points in AI development is going from idea → execution. 'cerebrium run' fixes that - execute code in the cloud in 1-2 seconds with no provisioning delays and no CI/CD. #startups #cloud #gpu #code #dev

Today, we’re trusted by teams building the future of AI such as @heytavus , @Vapi_AI , @DeepgramAI and many others. Full announcement here → We're hiring ->

We offer:.⚡️ Sub-2s cold starts.🌍 Multi-region support.📦 Elastic scaling.⏱️ Per-second billing.🔒 Enterprise-grade security & data residency. Zero DevOps. Just code.

Cerebrium abstracts the infrastructure layer — so you can go from prototype to production without hiring an infra team. Whether you’re:.• Deploying LLMs.• Running voice agents.• Doing batch data processing. —we make it fast, secure, and scalable.

Building real-time, multimodal AI apps is still way too hard. Teams wrestle with:.• GPU provisioning.• DevOps overhead.• Sky-high cloud bills. The result? Slow iteration cycles & a ton of wasted compute.

We’re excited to share that we’ve raised an $8.5M seed round to scale the high-performance, serverless infrastructure platform for AI. Led by @GradientVC, with participation from @ycombinator Authentic Ventures, and an incredible group of angels and operators. 🧵👇

Cost: ~$0.03/min per call. Production-ready. Globally distributed. Fully autoscaled. If you’re building an AI voice agent (or thinking about it), this is your blueprint. Blog: Github:

The voice stack we cover:.• STT (e.g. Deepgram, Whisper).• LLM (e.g. Llama 3 on vLLM or SGLang).• TTS (e.g. Rime Labs, Sesame).• Agent logic (LiveKit Agents).• Media transport (WebRTC).

🌍 Regional deployment is key. We show how deploying each component in the same Cerebrium region (US, EU, UAE, India) gives:.✅ <10ms inter-service latency.✅ Sub-500ms response times.✅ Data residency compliance (HIPAA, GDPR, etc.).

The key to hitting low latency?. Deploy everything inside the same cluster. No cross-region hops. No external API calls. Just local inference. Use open-source models like Llama or Sesame, or deploy partner models from @DeepgramAI (STT) and @rimelabs (TTS) directly on Cerebrium.

🚀 Deploying a real-time AI voice agent with ~500ms latency — globally. We just dropped a full breakdown of how to assemble a low-latency speech pipeline using STT, LLMs, and TTS, optimized for production use. Here’s how we did it with @livekit 🧵.

RT @garvinechan: deployed my voice ai agent on @cerebriumai . super quick and easy to use for deploying ai app. highly recommend!.

Our CEO, @MichaelLouis_za , shares some of the learnings today with the Pipecat community on how we have been able to get voice agents lower than 500ms end-to-end - highly recommend signing up for the course!.

Tutorial: Github:

⚡ Voice AI that responds in 600ms?. Meet Ultravox — a multimodal LLM from @FixieAI that skips the STT step and processes audio directly into an LLM. Built on @cerebriumai infra, deployed with Pipecat from @trydaily

RT @kwindla: $500 in Cerebrium credits (we're up to $5,500 in total credits so far) . Last June we worked with the team at @cerebriumai….

Check out the article and code here:. Article: Github: