Simuser Framework Engineer for AI Simulations

Kindredventures; San Francisco (CA); On-site; USD 175,000 - 250,000; Full time

Benefits offered by this job

In-office perks and daily lunch

Medical, dental, and vision benefits

Unlimited PTO

Equity stake

Job summary

Scorecard is seeking a Research Engineer to build a simuser orchestrator and a scalable simuser framework. You will convert real transcripts into digital twins that reflect user goals, knowledge and quirks, and you will design rigorous evaluations to quantify fidelity.

You will ship the framework so simusers operate inside training environments and contribute to measurable improvements in AI evaluation. You will work with production Python/TypeScript, iterate quickly, and own design decisions

Qualifications

• Proficient in production Python or TypeScript and able to own a system from design through deploy.
• Experience building agentic loops or multi-agent systems with eval in the loop.
• Fidelity-focused: design evaluations with baselines, held-out data, and calibrated judges.

Responsibilities

• Build the simuser orchestrator to manage conversations and context among simusers.
• Build the simuser framework with a structured schema for goals, knowledge, and style.
• Convert transcripts into simusers by turning work context and transcripts into a digital twin.
• Design and run evaluations to quantify fidelity of grounded simulated users.
• Ship the framework to production so simusers run inside training environments.

Skills

Python

TypeScript

LLM fluency

Multi-agent systems

Job description

Scorecard is seeking a Research Engineer to build a simuser orchestrator and a scalable simuser framework. You will convert real transcripts into digital twins that reflect user goals, knowledge and quirks, and you will design rigorous evaluations to quantify fidelity.

You will ship the framework so simusers operate inside training environments and contribute to measurable improvements in AI evaluation. You will work with production Python/TypeScript, iterate quickly, and own design decisions