Rugged, fanless, IP67-rated AI compute platforms engineered in Bengaluru for autonomous mobility, defence unmanned systems, and industrial OEM deployment. Built on NVIDIA Jetson and Qualcomm Robotics — supplied as production-ready boxes or as ODM reference platforms for your Tier-1 BoM.
From low-SWaP indoor AGVs to L4 autonomous trucks and humanoid platforms, Aumnium's Vajra and Garuda families cover the operational compute envelope end-to-end. Every unit ships with BSP, ROS 2 reference packages, and a 5-year supply commitment.
NVIDIA Jetson Orin™ Nano · 40 TOPS
Qualcomm® QRB5165 · 15 TOPS · Low SWaP
NVIDIA Jetson Orin™ NX · 100 TOPS
NVIDIA Jetson AGX Orin™ · 275 TOPS
Side-by-side technical comparison across the Aumnium edge AI compute family. Detailed mechanical, electrical, and software specifications are available in the per-product datasheets.
| // SPEC | VAJRA LITE | GARUDA Q8 | VAJRA NX | VAJRA AGX | VAJRA THOR |
|---|---|---|---|---|---|
| SoM / SoC | Orin Nano | QRB5165 | Orin NX | AGX Orin | Jetson Thor |
| AI Performance | 40 TOPS | 15 TOPS | 100 TOPS | 275 TOPS | 2000 TFLOPS FP4 |
| CPU | 6× A78AE | 1× Kryo 585 8-core | 8× A78AE | 12× A78AE | 14× Neoverse V3AE |
| Memory | 8 GB LPDDR5 | 8 GB LPDDR5 | 16 GB LPDDR5 | 32 GB LPDDR5 | 128 GB LPDDR5X |
| Storage (max) | 1 TB NVMe | 512 GB UFS | 2 TB NVMe | 4 TB NVMe | 8 TB NVMe |
| Camera Inputs | 4× MIPI / 2× GMSL2 | 7× MIPI ISP | 8× GMSL2 | 16× GMSL2 | 16× GMSL3 |
| Network | 1× GbE | 1× GbE + Wi-Fi 6 | 2× GbE PoE+ | 2× 10 GbE | 4× 25 GbE |
| Power Range | 7–15 W | 5–12 W | 10–25 W | 15–60 W | 40–130 W |
| DC Input | 9–36 VDC | 9–36 VDC | 9–48 VDC | 9–48 VDC | 9–60 VDC |
| Operating Temp | −25 → +70 °C | −30 → +75 °C | −40 → +75 °C | −40 → +75 °C | −40 → +85 °C |
| IP Rating | IP54 | IP65 | IP67 | IP67 | IP67 |
| MIL-STD-810G | — | ✓ shock | ✓ shock + vibe | ✓ full | ✓ full |
| ROS 2 BSP | ✓ Humble | ✓ Humble | ✓ Humble + Jazzy | ✓ Humble + Jazzy | ✓ Jazzy |
| Form Factor | 180×130×60 mm | 165×120×55 mm | 270×195×78 mm | 290×220×95 mm | 340×260×120 mm |
| Weight | 1.4 kg | 1.1 kg | 2.4 kg | 3.6 kg | 5.8 kg |
| Starting Price (USD) | $1,295 | $1,495 | $2,295 | $4,495 | $8,995 |
// All specifications subject to revision. Volume discounts apply at 25, 100, 500, and 1000+ unit tiers. Customizations available — see ODM section.
From contested-environment defence platforms to lithium-mine haulers and last-mile delivery rovers — every Aumnium SKU is qualified against a specific operational envelope. Pick the domain; we'll spec the box.
UGV, USV, ASV and tactical sensor mast compute. MIL-STD-810G shock and vibration; sealed against EMI/RFI ingress; secure boot and tamper-evident enclosure available.
L3/L4 perception stacks for haul trucks, last-mile delivery, autonomous taxis. 16-camera GMSL2 multi-stream support; 1588 PTP time sync; automotive-grade DC input range.
Low-SWaP edge compute for fixed-wing surveillance UAVs, multirotor inspection drones, and autonomous swarms. Garuda Q8 is purpose-built for sub-2 kg payload constraints.
Autonomous haul trucks, dragline monitoring, conveyor inspection. Wide DC input handles 24V/48V truck buses; survives diesel particulate, dust, and equatorial heat.
Indoor warehouse AGVs, hospital service robots, retail inventory bots, and humanoid platforms. Vajra Lite and Vajra NX cover the 95% of AMR compute envelopes.
Autonomous surface vessels, hull-inspection ROVs, port monitoring. IP67 rating with optional pressure-rated enclosures for shallow-water deployment.
Roadside ITS units, traffic-camera AI accelerators, smart-junction perception. PoE+ powered, fanless, designed for cabinet deployment with passive thermal handling.
Reference designs, schematic packages, and joint productisation programs. Bring us your form factor and we'll deliver a production-ready board with NVIDIA or Qualcomm silicon.
We are an India-headquartered edge AI engineering firm built to serve two customer types: defence integrators who need procurement-friendly domestic compute, and global OEMs who need a competent ODM partner with deep hardware-software vertical integration.
Designed, assembled, and tested in Bengaluru. ITAR-friendly supply chain. Domestic procurement-eligible for Indian MoD and MHA contracts under DAP-2020 categories.
Most competitors are mono-vendor. We carry credentials with both NVIDIA Jetson Ecosystem and Qualcomm Robotics — letting you spec the right silicon for your TOPS-per-watt curve, not your supplier's.
Every box ships with Aumnium-maintained BSP, GMSL camera drivers, ROS 2 Humble/Jazzy reference applications, and validated SLAM/perception model zoo.
Aerospace and defence programs need 5-7 year horizons. We commit BoM-frozen production with revision change control, lifetime-buy options, and last-time-buy notification 18 months before EoL.
MIL-STD-810G shock, vibration, salt-fog, and humidity-cycle testing performed in our own lab. Reports delivered with each unit. Optional EMI/RFI compliance to MIL-STD-461F.
Bring us your ICD; we'll deliver a custom board, mechanical, BSP, and certification package in 18-24 weeks. Joint IP terms or work-for-hire — whichever fits your program.
Tell us your application, target volume, and deployment timeline. Our applications-engineering team will respond within 1 business day with indicative pricing, datasheets, and an engineering call to scope customisations.