Spaces:
Configuration error
Configuration error
| title: Future Present Labs | |
| short_description: AI-forward manufacturing and open-source tools. | |
| # Future Present Labs | |
| Future Present Labs is an AI-forward manufacturing studio in Belltown, Seattle. | |
| We build software, machines, models, fixtures, and finished parts in the same | |
| loop, with a bias toward practical tools that make physical work faster, easier | |
| to inspect, and easier to automate. | |
| Our shop combines precision manufacturing with applied AI: | |
| - 4-axis CNC milling, CNC turning, waterjet cutting, laser marking, PCB design, | |
| 3D printing, mold-making, finishing, and assembly. | |
| - Agentic workflows for quoting, DFM review, routing, purchasing, inspection, | |
| documentation, and customer communication. | |
| - Open-source infrastructure for people and agents that need to move from CAD, | |
| BOMs, prompts, notes, or measurements to real-world parts. | |
| We use Hugging Face to share experiments, models, demos, datasets, and tooling | |
| from our manufacturing workbench. Some projects are small utilities; others are | |
| building blocks for larger systems we run internally at [fpl.dev](https://fpl.dev). | |
| ## What We Are Building | |
| **Transmog** explores how agents can transform messy product intent into | |
| manufacturable artifacts: CAD-adjacent data, files, quotes, routings, and | |
| handoffs that humans can still inspect. | |
| **Legion of BOM** is our work on bill-of-material intelligence: parsing, | |
| normalizing, enriching, comparing, and reasoning over BOMs so hardware projects | |
| can move with less spreadsheet archaeology. | |
| **JARVIS voice work** is focused on a practical shop-floor assistant: voice | |
| interfaces for machines, job status, documentation, reminders, and engineering | |
| context, tuned for real work rather than demos alone. | |
| Alongside those projects, we are developing smaller AI systems for supplier | |
| research, quoting, scheduling, inspection notes, CAD review, customer support, | |
| and internal operations. | |
| ## How We Think About AI and Manufacturing | |
| Manufacturing is full of narrow, high-context decisions. A useful AI system in | |
| this environment needs to respect tolerances, materials, machine constraints, | |
| revision history, lead times, operator judgment, and the cost of being wrong. | |
| Our approach is human-in-the-loop by default. Agents should prepare, check, | |
| summarize, route, and accelerate work, but the final loop still belongs to the | |
| people responsible for the part, the process, and the customer. | |
| We care about systems that are: | |
| - **Grounded:** connected to real files, real machines, real measurements, and | |
| real constraints. | |
| - **Auditable:** easy to inspect, replay, correct, and improve. | |
| - **Composable:** useful as small pieces, not just as one big application. | |
| - **Open where possible:** shared when it helps others build, learn, or verify. | |
| ## Open Source | |
| We like open-source because it is how good technical ideas get pressure-tested. | |
| When we can publish code, models, prompts, datasets, specs, or demos without | |
| violating customer trust or operational security, we try to do that. | |
| Not everything from a working machine shop can be public, but our goal is to | |
| share useful pieces of the stack as they become stable enough for others to run, | |
| fork, criticize, and improve. | |
| ## Links | |
| - Website: [fpl.dev](https://fpl.dev) | |
| - Hugging Face: [FuturePresentLabs](https://huggingface.co/FuturePresentLabs) | |