{"id":101528,"title":"Zibra Labs: HPC for Quant Trading Firms Running Backtesting at Scale","tagline":"HPC Clusters for Parallel Workloads on Spot-instances Across Hyperscalers and Neoclouds.","body":"### Credentials\n\nWe’ve spent years working together building large scale distributed systems that run on hundreds of thousands of machines and serve billions of users. **Every time you leverage your network for a warm intro on LinkedIn, you use three databases that we’ve built: [Venice](https://venicedb.org/),[ Liquid](https://www.linkedin.com/blog/engineering/graph-systems/liquid-the-soul-of-a-new-graph-database-part-1), and[ Espresso](https://engineering.linkedin.com/teams/data/data-infrastructure/storage-infra/espresso).**\n\nAfter LinkedIn, **we were tech-leads of [Ray](https://github.com/ray-project/ray)**, the leading open-source compute platform used by companies such as xAI, Cursor, Bridgewater, P72, Man Group, Two Sigma, and others. Ray has [\\~12M](https://pypistats.org/packages/ray) weekly downloads and is part of the PyTorch Foundation.\n\n### Product\n\n\u003chttps://www.youtube.com/watch?v=C6tl8Fr8t2I\u003e\n\nWe’re building large scale high performance computing (HPC) clusters for **quantitative trading firms** to run **parallel simulation workloads such as backtesting.** \n\nThe scale we’re targeting is\n\n* 100 to 50,000 nodes of mixed hardware (CPUs and GPUs) on hyperscalers and neoclouds.\n* up to 5,000,000 parallel tasks.\n* spot instances for cheaper compute across multiple availability zones, hyperscalers, and neoclouds.\n* \u0026lt; 50ms overhead for task dispatch and scheduling.\n* run the compute in your cloud or on-prem environment to keep your alpha secure.\n\nOur goal is to **make testing trading strategies fast, cheap, and reliable.** \n\n### Vision\n\nWe know that existing solutions do not handle this scale, do not have first-class support for spot-instances, and will not match our low-latency and high throughput performance. \n\nOur technology generalizes to other large-scale compute workloads such as\n\n1. **post-training and reinforcement learning (RL)** for frontier and applied AI labs\n2. **massively parallel robotics simulation and sim-to-real training** for robot foundation model teams\n3. **virtual screening and molecular dynamics** for techbio and pharma R\u0026amp;D\n4. **scientific simulation** (CFD, climate, seismic, FEA) for HPC teams looking to move beyond MPI and SLURM.\n\n### Ask\n\nReach out if you\n\n1. Are running simulation workloads that require lots of computation at a quant trading firm.\n2. If you need HPC for a different use-case.\n3. Want to grab a beer and talk about distributed systems or HPC.\n\nCheers!\n\nIbrahim and Zac","slug":"QPY-zibra-labs-hpc-for-quant-trading-firms-running-backtesting-at-scale","created_at":"2026-05-18T20:53:24.005Z","updated_at":"2026-05-25T04:24:16.318Z","total_vote_count":5,"url":"https://www.ycombinator.com/launches/QPY-zibra-labs-hpc-for-quant-trading-firms-running-backtesting-at-scale","share_image_url":"//bookface-static.ycombinator.com/assets/ycdc/yc-og-image-c440a0ad1dacfb86eeeb343717479cc54d256614449b4ef719977a0a451f8bc8.png","company":{"id":31459,"name":"Zibra Labs","slug":"zibra-labs","url":"https://zibralabs.ai/","logo":"https://bookface-images.s3.amazonaws.com/small_logos/561de55fc1ba8b5c7254d0ad9eed2d656139b041.png","batch":"Spring 2026","industry":"B2B","tags":["Artificial Intelligence","Fintech","Cloud Computing","Infrastructure"],"search_path":"https://bookface.ycombinator.com/company/31459"}}