Skip to main content
TALENT247 // SPECIALIST AI EXECUTIVE SEARCH & TECHNICAL RECRUITMENT
APP · Applied AI

ML Solutions & Forward-Deployed Inference Engineer

San Francisco, CA · Full-time · Hybrid

AI Specialism

Applied AI

Workplace

Hybrid

Location

San Francisco, CA

Compensation

$190k – $265k

annual package

Role Dossier & Specification

REF: ML-SOLUTIONS-FORWA

THE CLIENT & MANDATE

Our client is a premier San Francisco AI inference platform trusted by category-defining voice AI, coding assistant, healthcare, and multimodal media companies to run custom models in production at massive scale. Their ML Solutions & Forward-Deployed Inference team acts as the tip of the spear—partnering directly with founders and Staff ML engineers at breakout AI companies to take custom PyTorch checkpoints from research notebooks to sub-100ms global production endpoints.

They have engaged Talent247 to recruit an ML Solutions & Forward-Deployed Inference Engineer in San Francisco, CA. Blending hands-on GPU model optimization with high-velocity customer architecture, the appointed engineer will solve gnarly real-world serving bottlenecks across LLMs, streaming audio, vision-language, and diffusion pipelines.

KEY RESPONSIBILITIES

  • Partner directly with customer ML engineering teams during critical launches to package, benchmark, quantize, and deploy custom PyTorch, TensorRT-LLM, SGLang, and vLLM pipelines onto dedicated GPU clusters.
  • Profile end-to-end request lifecycles—from tokenization, audio/image preprocessing, and dynamic batching through GPU kernel execution and WebSocket/gRPC streaming—cutting latency by 2x–5x and GPU cost by 40%+.
  • Implement production-ready compound inference chains combining custom LoRA adapters, Whisper/streaming ASR, expressive TTS, embeddings/rerankers, and speculative decoding draft models.
  • Design realistic load-testing suites matching customer burst concurrency, context-length distributions, and multi-turn prefix-cache patterns prior to production cutover.
  • Build clean open-source reference templates, Python packaging SDK improvements, and automated quantization recipes benefiting thousands of platform developers.
  • Serve as the voice of frontier ML builders internally, shaping product roadmaps across core model performance and multi-cloud infrastructure teams.

TECHNICAL ENVIRONMENT & OFFERING

Stack: Python, PyTorch, TensorRT-LLM, vLLM, SGLang, Triton Inference Server, Docker/Kubernetes, FastAPI/gRPC/WebSockets, FP8/INT4 quantization toolkits, H100/H200/B200 GPUs. Workplace: Hybrid in San Francisco, CA. Compensation: $190,000 – $265,000 base salary plus meaningful equity grants, performance incentives, and full US benefits.

Core Requirements

ESSENTIAL QUALIFICATIONS

  • 4+ years of ML engineering, inference optimization, or forward-deployed solutions engineering experience deploying deep learning models into latency-sensitive production environments.
  • Strong hands-on proficiency in PyTorch alongside modern LLM/generative serving engines (vLLM, TensorRT-LLM, SGLang, or Diffusers/TorchCompile).
  • Practical command of GPU memory sizing, KV-cache math, continuous batching, tensor parallelism, and quantization trade-offs across L40S, H100, H200, and B200 GPUs.
  • Crisp Python systems engineering skills (asyncio, multiprocessing, shared memory, containerization, OpenTelemetry/Prometheus metrics).
  • Articulate, low-ego collaborator energized by debugging live technical challenges alongside elite AI engineering teams.

PREFERRED / BONUS EXPERIENCE

  • Experience optimizing real-time streaming audio/voice pipelines (TTS/ASR), video/image diffusion models, or custom C++/CUDA PyTorch extensions.
  • Prior forward-deployed, applied AI, or developer-facing ML engineering experience at a high-growth AI infrastructure company.
  • Published technical benchmarks, blog posts, or open-source inference recipes.

Position Telemetry

Location San Francisco, CA (Hybrid)
Department Applied AI
Employment type Full-time
Workplace type Hybrid
Salary range
$190k – $265k annual package