Staff AI Architect: Agents & Model Adaptation

People in AI; New York (NY); On-site; USD 225,000 - 275,000; Full time

Job summary

People in AI in New York City seeks a Staff Applied AI/ML Engineer to shape agent architecture and model strategy while remaining hands-on. You’ll decide when frontier models or adapted models are best and prove approaches in production.

You’ll design agent systems, tune models, build evaluation harnesses, and mentor engineers, delivering reliable, scalable AI that handles messy data and real-world consequences.

Qualifications

• Experience shipping production AI systems.
• Hands-on LLM fine-tuning with clear training choices and results.
• Ability to own technical direction and mentor engineers.

Responsibilities

• Design and build agent systems spanning tool use, context management, orchestration, and guardrails.
• Choose models based on accuracy, latency, cost, and operational requirements.
• Fine-tune open-weight models using LoRA, QLoRA, supervised fine-tuning, and preference optimization.
• Turn user corrections and traces into training data and evaluation cases.
• Build evaluation harnesses measuring accuracy, hallucinations, false positives, and task completion.
• Own production serving decisions including batching, quantization, monitoring, and inference efficiency.
• Test systems against prompt injection, data leakage, and adversarial inputs.
• Guide architecture, review designs, and mentor engineers while continuing to write code.

Skills

Software engineering

Agent systems

LLM fine-tuning

Evaluation design

Architecture direction

Mentoring engineers

Reliability/security/cost

Tools

Python

PyTorch

Job description

People in AI in New York City seeks a Staff Applied AI/ML Engineer to shape agent architecture and model strategy while remaining hands-on. You’ll decide when frontier models or adapted models are best and prove approaches in production.

You’ll design agent systems, tune models, build evaluation harnesses, and mentor engineers, delivering reliable, scalable AI that handles messy data and real-world consequences.