Software Engineer

Key Talent Solutions; San Francisco (CA); On-site; USD 250,000 - 350,000; Full time

Benefits offered by this job

Profit sharing

Equity

Job summary

Key Talent Solutions leads a role with an AI startup focusing on datasets and evaluation systems used to train frontier models. You collaborate with top research teams to define tasks and the scoring systems that guide model improvement.

The role requires hands-on ML/LLM experience, sharp data quality instincts, and 1–4 years shipping technical ML work. Startup founder experience is advantageous, with a base salary of $250k plus profit share and equity.

Qualifications

• Hands-on with ML/LLM workflows training, fine-tuning or evaluation.
• Sharp instincts for data quality, edge cases and precision/recall trade-offs.
• 1–4 years shipping technical work in ML/AI (RLHF/RLVR or eval).

Responsibilities

• Design data that exposes where models fail across finance, code & enterprise workflows.
• Build reward signals and evaluation rubrics for RLHF/RLVR training pipelines.
• Develop frameworks to measure dataset quality and its impact on model performance.
• Turn ambiguous research goals into concrete, shippable systems.

Skills

ML/LLM workflows

Data quality

RLHF/RLVR

Startup background

Education

BS in CS/ML

Tools

Python

PyTorch

TensorFlow

Job description

We're partnering with a fast-growing (Series B) AI company that builds the datasets and evaluation systems frontier AI labs use to train their models. You'd be working hand-in-hand with top research teams designing the tasks models practise on, and the scoring that decides whether they're really getting smarter.

What you'd own:

• Design data that exposes where models fail across finance, code & enterprise workflows
• Build reward signals and evaluation rubrics for RLHF / RLVR training pipelines
• Develop frameworks to measure dataset quality and its real impact on model performance
• Turn ambiguous research goals into concrete, shippable systems

You must have:

• Hands-on with modern ML/LLM workflows training, fine-tuning or evaluation
• Sharp instincts for data quality, edge cases and precision/recall trade-offs
• 1–4 years shipping technical work (RLHF/RLVR or eval experience)
• Previous experience as a startup founder or co-founder is advantageous

The details:

$250 base + significant profit share + equity

This is a rare chance to have direct, measurable impact on frontier AI on a small team where your work ships straight into the models defining the field.