Fractile Tech Stack

AI inference accelerator chips and systems for data-center deployment

Computer Hardware Manufacturing London, London 51–200 employees Founded 2022 Privately Held

Fractile designs custom silicon and systems software for AI inference workloads, with a stack spanning hardware design (SystemVerilog, Cadence Innovus, Synopsys ICC2), ML frameworks (PyTorch, vLLM, SGLang), and system-level tooling (Kubernetes, Linux kernel drivers in Rust). The engineering-dominated hiring mix (33 of 35 roles) and active projects spanning pre-silicon modeling through production deployment reveal a company moving from design validation toward real data-center inference integrations.

Tech Stack 42 technologies

Core StackGitHub Actions Python Rust Linux Docker Kubernetes PyTorch Bazel C C++23 SystemVerilog Cocotb Verilator ARM x86-64 RISC-V UVM SystemC MLIR C/C++ vLLM SGLang QEMU Linux Kernel Apple Silicon Cadence Innovus Synopsys ICC2 Calibre TCL Perl+12 more

AdoptingRust

What Fractile Is Building

◆Challenges

Launching ai accelerators into data centre deployment
Accelerating large language models
Ensuring manufacturability
Meeting demand and scaling product delivery
Inference deployment challenges
Scaling ci configuration
Scaling bazel builds
Running large language models 100x faster
Increase tape-out quality
Yield ramp acceleration

▲Active Projects

Launch ai accelerators into data centre deployment
Develop systems software for ai accelerator hardware
Pre-silicon modelling of ai accelerator hardware
Early software development using models of products in development
Runtime stack for ai accelerators
Ai accelerator deployment
Ml/ai ic product for scale-out to data centre
System-level hardware defect detection tests
Linux kernel driver in rust
Inference server integrations

Hiring Activity

Decelerating35 roles · 8 in 30d

Department

Engineering

Manufacturing

Seniority

Senior

Lead

Mid

Junior

Staff

Company intelligence

Find more companies like Fractile by tech stack, pain points and active projects

Get started free

About Fractile

Fractile, founded in 2022 and based in London, builds custom AI accelerator hardware and software stacks targeting inference at scale. The company addresses a specific gap in the AI compute market: while existing hardware excels at training large models, inference—the dominant operational workload—remains inefficient and costly. Fractile's approach rearchitects the memory-compute boundary through custom silicon, paired with a runtime stack and ML framework integrations. Active projects span pre-silicon validation, Linux kernel drivers, and inference server deployments. The organization is primarily engineering-focused, with hiring concentrated in the UK and Taiwan.

HeadquartersLondon, London

Company Size51–200 employees

Founded2022

Hiring MarketsUnited Kingdom, Taiwan

Frequently Asked Questions

What is Fractile building?

Custom AI accelerator chips and systems software optimized for inference workloads at scale. The company designs silicon (SystemVerilog, Cadence/Synopsys CAD tools) integrated with ML runtimes (vLLM, SGLang) and Linux kernel drivers (Rust).

What tech stack does Fractile use?

Hardware design: SystemVerilog, Cadence Innovus, Synopsys ICC2, Calibre. ML/software: PyTorch, vLLM, SGLang, MLIR. Systems: Kubernetes, Docker, Linux kernel, Rust. Build: Bazel, GitHub Actions. Simulation: Verilator, Cocotb, QEMU.