# Type 1 Compute

Type 1 Compute converts existing AI models (PyTorch, ONNX) to run on FPGAs with 10x better power efficiency than Nvidia Jetson Nano. No retraining required.

## What we do
- FPGA inference optimization via model conversion
- Managed service — we handle conversion, deployment, and maintenance
- Custom ASIC development roadmap (100x efficiency target)

## Benchmarks
- 75.76 GOP/s/W (FPGA prototype, DVS128 gesture recognition dataset)
- 9.5x more efficient than Jetson Nano
- 244x more efficient than Intel i9 CPU
- 5.7x efficiency gain on event-based object detection (SpikeYOLO)
- <1W power consumption for UAV edge inference
- 0.8ms latency for radar classification

## Best for
- Platforms with hard SWaP constraints
- GPS-denied or comms-denied environments
- Teams already running FPGAs
- Sparse, event-driven sensor workloads: radar, RF, acoustic, lidar
- Defense and aerospace applications requiring radiation tolerance
- Autonomous systems with strict power budgets

## Sectors
- Defense (Active)
- Telecom (Accepting Partners)
- Industrial (Accepting Partners)
- Medical (Research Stage)

## Open source
- https://github.com/type1compute/SpikeYoloV8-Tracker
- https://github.com/type1compute/SPIKE-SPACE-T1C
- https://github.com/type1compute/Spectrum-Analyzer

## Backed by
- Israel Aerospace Industries
- Stanford StartX
- Antler VC
- Mana Ventures
- Ollin Ventures
- Gaingels
- Hulsey Richmond Ventures
- NVIDIA Inception Program

## Academic partners
- Stanford University
- SLAC National Accelerator Laboratory
- University of Southern California
- UC Santa Cruz
- University of Milano-Bicocca
- Istinye University

## Contact
support@type1compute.com
https://www.type1compute.com