AI Architect for Software Development

Lattice Semiconductor • San Jose, CA

3d • $260,000 - $280,000

About The Position

We are looking for an AI Architect to design and lead the development of a large‑scale Agentic AI system that automates complex engineering workflows using LLMs, VLMs, and structured reasoning. You will architect multi‑agent workflows, build high‑quality training datasets, fine‑tune and optimize frontier models, and design the guardrails, retrieval pipelines, and evaluation frameworks required to run AI safely and reliably in production. This role is ideal for someone who excels in AI systems architecture, agent design, LLM/RAG pipelines, and model optimization.

Requirements

• 5+ years building ML or LLM‑powered products (multi‑agent systems preferred).
• Experience with LLM fine‑tuning (LoRA/QLoRA), distillation, and prompt engineering.
• Strong understanding of LLM internal architecture (transformers, attention, tokenization).
• Expertise in Python, PyTorch, and modern AI frameworks (HuggingFace, vLLM, LangChain).
• Experience building robust RAG pipelines (embedding optimization, retrieval metrics).
• Experience designing structured model interactions (JSON schema, function‑calling, tool‑use).

Nice To Haves

• Experience with embedded, hardware, or FPGA workflows—but not mandatory.

Responsibilities

• Design multi‑agent workflows with an orchestrator, specialized agents, and shared memory.
• Define agent interfaces, policies, and error‑handling/repair strategies.
• Implement strict schema‑bound LLM calls (JSON, XML, structured tool commands).
• Select, distill, fine‑tune, and evaluate LLM/SLM/VLA models.
• Design prompt templates, adapters, LoRA modules, and controlled‑generation methods.
• Build document ingestion pipelines with chunking, metadata tagging, and embeddings.
• Build domain‑aware RAG systems with citation‑based referencing.
• Develop multi layer guardrails using schema validation, rule based checks, and policy engines.
• Implement self critique, repair loops, and fallback strategies for model outputs.
• Build model serving pipelines (vLLM/TGI/Triton), caching, batching, and streaming.
• Implement logging, tracing, observability, and offline evaluation pipelines.

Benefits

• In addition to base salary, we offer an incentive plan bonus, and new hire equity for a competitive total compensation package.
• healthcare and retirement plans, paid time off, and more!

What This Job Offers

• Job Type: Full-time
• Career Level: Senior
• Education Level: No Education Listed