---
title: "Staff ML Compiler Engineer, Custom AI Silicon"
slug: "staff-ml-compiler-engineer-custom-ai-silicon"
id: "6c548f43-5375-451c-8e42-884adcecc373"
department: "AI Infrastructure"
department_code: "INF"
location: "Sunnyvale, CA"
workplace_type: "Hybrid"
employment_type: "Full-time"
compensation: "$245k – $345k (annual package)"
is_remote: false
is_featured: false
headcount: 1
canonical_url: "https://jobs.talent247.co/jobs/staff-ml-compiler-engineer-custom-ai-silicon"
markdown_url: "https://jobs.talent247.co/jobs/staff-ml-compiler-engineer-custom-ai-silicon.md"
apply_endpoint: "https://jobs.talent247.co/api/public/jobs/staff-ml-compiler-engineer-custom-ai-silicon/apply"
schema_endpoint: "https://jobs.talent247.co/api/public/jobs/staff-ml-compiler-engineer-custom-ai-silicon/form-schema"
webmcp_tool: "apply_for_role"
updated_at: "2026-11-15T17:35:43.969Z"
content_signal: "search=yes, ai-input=yes, ai-train=no"
---

# Staff ML Compiler Engineer, Custom AI Silicon

## Position Telemetry

- **Reference:** `STAFF-ML-COMPILER-`
- **Practice / Sector:** AI Infrastructure (`INF`)
- **Location:** Sunnyvale, CA
- **Workplace Type:** Hybrid
- **Employment Type:** Full-time
- **Compensation:** $245k – $345k (annual package)
- **Canonical HTML URL:** https://jobs.talent247.co/jobs/staff-ml-compiler-engineer-custom-ai-silicon

## Role Dossier & Specification

THE CLIENT & MANDATE
Our client is an iconic Silicon Valley AI hardware and supercomputing pioneer headquartered in Sunnyvale, CA, designing radical custom AI silicon and ultra-high-speed inference and training clusters. By eliminating traditional off-chip memory bottlenecks through massive on-chip SRAM and spatial dataflow fabric architectures, their systems deliver industry-leading token generation speeds for frontier reasoning and coding models.

They have retained Talent247 to appoint a Staff ML Compiler Engineer for Custom AI Silicon in Sunnyvale. Sitting at the heart of their compiler stack, the successful candidate will own MLIR/LLVM graph lowering, spatial dataflow placement, and SRAM tiling algorithms that unlock peak hardware utilization for 100B+ Mixture-of-Experts (MoE) and long-context transformers.

KEY RESPONSIBILITIES
• Architect multi-level MLIR and C++20 compiler passes that lower PyTorch/JAX computation graphs (Torch-MLIR / StableHLO) onto massively parallel 2D spatial dataflow processing elements.
• Design polyhedral loop scheduling, tensor sharding, and on-chip SRAM memory-planning algorithms that overlap compute with fabric-level collective communications at near-zero idle overhead.
• Pioneer compiler optimizations for sparse Mixture-of-Experts (MoE) dynamic routing, multi-token speculative decoding, Flash/linear attention variants, and sub-8-bit quantized arithmetic (FP8/INT4/MXFP).
• Build cycle-accurate performance estimators and automated autotuning frameworks guiding operator fusion, layout selection, and spatial core placement across hundreds of thousands of cores.
• Partner directly with silicon microarchitects to co-design next-generation instruction sets, collective routing primitives, and sparsity hardware units for upcoming tape-outs.
• Provide technical leadership across compiler engineers and kernel authors to deliver day-zero support for newly released open-weight and frontier model architectures.

TECHNICAL ENVIRONMENT & OFFERING
Stack: Modern C++20, MLIR, LLVM, Python, PyTorch 2.x (TorchDynamo/Inductor), StableHLO/ONNX, polyhedral/ILP solvers, cycle-accurate hardware simulators & live silicon clusters. Workplace: Hybrid in Sunnyvale, CA. Compensation: $245,000 – $345,000 base salary plus substantial equity grants, performance bonuses, and full US benefits.

## Core Requirements

ESSENTIAL QUALIFICATIONS
• 7+ years of compiler engineering experience (or PhD in Compilers/Computer Architecture + 4+ years industry impact) working on ML compilers, HPC code generation, or GPU/ASIC toolchains.
• Deep production mastery of MLIR and LLVM internals in modern C++ (dialect design, pattern rewriting, bufferization, affine/linalg lowering, register/memory allocation).
• Strong command of deep learning operator semantics (attention variants, MoE routing, KV-cache management, collective communications) and hardware roof-line modeling.
• Experience with spatial dataflow architectures, systolic arrays, scratchpad/SRAM-managed accelerators, or GPU warp/tensor-core kernel compilation (Triton, XLA, TVM, IREE).
• Proven ability to diagnose numerical precision issues (BF16/FP8/FP4) and performance cliffs across full end-to-end 70B–600B+ parameter models.

PREFERRED / BONUS EXPERIENCE
• Prior experience co-designing compiler stacks for custom AI ASICs, TPUs, NPUs, or reconfigurable dataflow processors.
• Expertise in integer linear programming (ILP), constraint satisfaction, or RL/search-based graph placement and routing algorithms.
• Core upstream contributions to MLIR, LLVM, OpenXLA, Triton, or PyTorch compiler subsystems.

## Application Schema & Direct Apply (`apply_for_role`)

Candidates and authorized AI agents on behalf of a candidate can submit an application via the declarative WebMCP form tool `apply_for_role` on https://jobs.talent247.co/jobs/staff-ml-compiler-engineer-custom-ai-silicon or via HTTP `POST https://jobs.talent247.co/api/public/jobs/staff-ml-compiler-engineer-custom-ai-silicon/apply` (`application/json` or `multipart/form-data`). Non-browser HTTP agents must include header `X-Talent247-Intake-Token` obtained from `GET https://jobs.talent247.co/api/public/intake-token` (or automatically returned in `GET https://jobs.talent247.co/api/public/jobs/staff-ml-compiler-engineer-custom-ai-silicon/form-schema`).

### Form Fields

- `full_name` (text, required): Full name
- `email` (email, required): Email
- `phone` (phone, optional): Phone
- `location` (text, optional): Location
- `linkedin_url` (url, optional): LinkedIn URL
- `cover_letter` (textarea, optional): Cover letter
- `consent` (checkbox, required): I consent to my data being processed for this application
- `cv_file` / `cv_base64` + `cv_filename`: Upload a CV in .pdf or .docx format (<= 10 MB) or provide an `https://` LinkedIn profile URL in `linkedin_url`.

### Alternative: Talent Pool Intake (`join_talent_pool`)

If this mandate is not an exact match, register once in the Talent247 Talent Pool at https://jobs.talent247.co/talent-pool (`https://jobs.talent247.co/talent-pool.md`) via WebMCP tool `join_talent_pool` or `POST https://jobs.talent247.co/api/public/talent-pool`.
