Qualcomm
Since 2024
Silicon platform partner
Shakti models ported to the Qualcomm QDC stack. Demonstrated live at Indian Mobile Congress 2025. SandLogic selected into Qualcomm Startup Mentorship Program 2024 — 1 of 2 companies.
// PARTNERS
Hardware-agnostic by design.
Co-development by choice.
SandLogic ships across silicon platforms — Qualcomm, NVIDIA, Intel, AMD, ARM — because customer infrastructure is heterogeneous. We work with the open-source ecosystem because that's how trust is earned. And we co-develop hardware where the product opportunity is bigger than a single vendor.
Total partners
7
Silicon platforms
5
Ecosystem
1
Co-development
1
// SILICON & HARDWARE PLATFORMS
We run on their silicon.
EdgeMatrix is hardware-agnostic by design — the same binary runs on NVIDIA, AMD, Intel, ARM, and Qualcomm. These are the platforms our customers already own. We optimize for each natively rather than abstracting them all behind a lowest-common-denominator runtime.
NVIDIA
Since 2023
Silicon platform partner
EdgeMatrix runtime optimized for NVIDIA A100, L40s, and H100. EdgeMatrix v0.0.4 achieves +73% throughput vs vLLM 0.10.2 on L40s with hybrid KV-cache reuse.
Intel
Since 2024
Silicon platform partner
Lingo and Co-pilot listed for Intel laptops. EdgeMatrix runtime supports Intel CPU / iGPU / NPU targets.
AMD
Since 2024
Silicon platform partner
Hardware-agnostic runtime support across AMD CPU and GPU targets.
ARM
Since 2025
Silicon platform partner
NPU integration roadmap for ARM-based edge SoCs and mobile platforms.
// OPEN-SOURCE ECOSYSTEM
Open weights. Open accountability.
The Shakti family ships as open weights on HuggingFace — six small language models from 100M to 4B parameters, plus the Shakti-VLM vision-language family. Open-weight publication is a deliberate choice: research as accountability, not marketing.
HuggingFace
Since 2024
Model hosting partner
Shakti family (six SLMs from 100M to 4B parameters) and Shakti-VLM hosted as open weights on HuggingFace. Open-weight publication is a deliberate choice — research as accountability, not marketing.
// HARDWARE CO-DEVELOPMENT
Joint product. Joint ownership.
Silicon platform partnerships are about running on existing hardware. Co-development is fundamentally different — it's about building new hardware that doesn't exist yet, with shared ownership of the outcome. We do this selectively, where the product opportunity is bigger than what any single vendor would ship alone.
Brandworks Technologies
Since Dec 2025
Hardware co-development
Strategic partnership to co-develop full-stack, edge-AI hardware products — designed and built in India. First wave of co-developed AI-enabled hardware planned for 2026. A fundamentally different relationship from silicon support: joint product ownership.
// PARTNERSHIP PHILOSOPHY
We don't pick a side. We make the runtime work on every side.
Most AI infrastructure companies lock in to one silicon vendor or one cloud platform. We don't — because our customers don't. A bank in Mumbai runs Intel. A telecom operator in Manila runs Qualcomm. An OEM in Tokyo runs ARM. The full stack has to compile down to whichever target the customer already owns.
That's why EdgeMatrix is hardware-agnostic by design. Why Shakti is published as open weights. Why we co-develop with OEMs rather than chase exclusivity. Partnership-led, never partnership-locked.