SubQ · the platform and cloud behind the research

SubQ (Subquadratic Inc.) is a frontier AI research and infrastructure company building subquadratic-attention LLMs. This page carries only public company information and role.

Company · Role · ProductsCaptured Jul 2026 · subq.ai

Case sheetFile 01 / 04 · SubQ (Subquadratic)

Role

Platform & Infrastructure Engineer

Timeline

2025 to present

Contribution split · I implemented

60%

infrastructure

40%

platform

cloud estate · developer platform

StackShow 7 items

TypeScript · Node.js · AWS · CDK · React · Serverless + containers · Managed data stores · OpenTelemetry

Linkssubq.ai ↗Technical report (PDF) ↗

Public information · sourced from subq.ai

Company

Subquadratic Inc. is a frontier AI research and infrastructure company building subquadratic-attention LLMs, a linear-time approach to long-context language modeling. Its tagline is Efficiency is intelligence. Two commercial products sit on top of the research: SubQ API (full-context inference for developers and enterprise teams) and SubQ Code (a long-context layer for Claude Code, Codex, and Cursor).

Everything in this section is drawn from subq.ai and the linked technical report.

Fig. 01 — SubQ · brand markFrom subq.ai

Where the work sits

Role

I joined SubQ in 2025 as platform and infrastructure engineer. The work sits at roughly a 60/40 split between infrastructure and platform: the cloud estate that everything runs on, and the developer-facing platform layered on top.

Infrastructure · ~60%

cloud estate

I design, build, and maintain the cloud infrastructure the developer platform runs on.

Platform · ~40%

developer platform

I design, build, and maintain the developer platform itself.

Two commercial products

What SubQ ships publicly

Product 01

SubQ API

The full-context API for developers and enterprise teams. Process full repositories and pipeline states in a single API call at linear cost. Public claims include a 12M-token context window and materially less compute than dense attention at long contexts.

Product page · subq.ai ↗

Product 02

SubQ Code

The long-context layer for coding agents. Plugs into Claude Code, Codex, and Cursor to map codebases, gather context, and answer token-heavy questions faster.

Product page · subq.ai ↗

Fig. 02 — Compute comparison · dense O(n²) vs SSA O(n)From subq.ai

Compute comparison chart from subq.ai: dense attention O(n²) versus SSA O(n), PFLOP per attention layer, showing an increasing multiplier at longer contexts

Chart published by Subquadratic on subq.ai · reproduced here as company context.

For the architecture and benchmarks behind these products, see the SubQ 1.1 Small model card and the posts on the SubQ blog.