AI/ML Engineer

Harnham; San Francisco (CA); On-site; USD 150,000 - 230,000; Full time

Job summary

Harnham is partnering with a high-growth startup building AI-native data infrastructure for Fortune 500 enterprises. The role focuses on owning the intelligence layer of the platform, making architectural decisions around agent design, model selection, and evaluation strategy.

Everything ships to customers; not just POCs or internal tooling. You'll own end-to-end AI agent architecture, including RAG pipelines, vector retrieval, and knowledge graphs, with emphasis on production quality,

Qualifications

• 4-6+ years of software engineering experience with meaningful AI/ML depth.
• Proven experience building and shipping AI agents or coding agents in production.
• Strong Python skills with hands-on experience across RAG, vector databases, and knowledge graphs.
• Experience with LLM APIs, prompt engineering, fine-tuning, and structured outputs in production.
• Familiarity with AWS and Kubernetes for production AI deployment.
• Experience with data platforms such as Databricks, Snowflake, BigQuery, or Spark is a strong bonus.
• Scale or Go experience is a bonus.

Responsibilities

• Architect multi-step AI agents end-to-end, covering problem decomposition, context retrieval, error recovery, and output quality.
• Own RAG pipelines, vector and graph-based retrieval systems, and knowledge representations that let agents reason accurately over complex enterprise data.
• Build the AI layer behind the platform's core code generation capability, from prompt design and model selection through to fine-tuning and output validation.
• Make production AI work at scale, balancing latency, throughput, and observability, and knowing when to optimize a model versus restructure a system.
• Think in failure modes and evaluation strategy, not capability demos, and own the improvement loop.

Skills

Python

AI/ML depth

RAG pipelines

LLM APIs

Prompt engineering

AWS

Kubernetes

Databricks/Snowflake/BigQuery

Tools

Vector databases

Knowledge graphs

Job description

We're partnered with a well-funded, high-growth Series B startup building the next generation of AI-native data infrastructure for Fortune 500 enterprises. With 3.5x revenue growth in FY'24 and 160% net revenue retention, they're scaling fast and investing heavily in the AI engineering team that powers their core product.

This is a high-impact role where you'll own the intelligence layer of the platform, making real architectural decisions around agent design, model selection, and evaluation strategy. Everything you build ships directly to customers; this is not a POC or internal tooling team.

What You'll Do

• Architect multi-step AI agents end-to-end, covering problem decomposition, context retrieval, error recovery, and output quality
• Own RAG pipelines, vector and graph-based retrieval systems, and knowledge representations that let agents reason accurately over complex enterprise data
• Build the AI layer behind the platform's core code generation capability, from prompt design and model selection through to fine-tuning and output validation
• Make production AI work at scale, balancing latency, throughput, and observability, and knowing when to optimize a model versus restructure a system
• Think in failure modes and evaluation strategy, not capability demos, and own the improvement loop

Requirements

• 4-6+ years of software engineering experience with meaningful AI/ML depth
• Proven experience building and shipping AI agents or coding agents in production
• Strong Python skills with hands-on experience across RAG, vector databases, and knowledge graphs
• Experience with LLM APIs, prompt engineering, fine-tuning, and structured outputs in production
• Familiarity with AWS and Kubernetes for production AI deployment
• Experience with data platforms such as Databricks, Snowflake, BigQuery, or Spark is a strong bonus
• Scale or Go experience is a bonus

If you're a builder who thinks in failure modes, owns what you ship, and thrives in environments where last year's best practice is already obsolete, this is a rare opportunity to have direct leverage on a product used by some of the world's largest enterprises.