Pantomath is building the autonomous layer for enterprise Data Operations. Its AI agents catch data incidents, trace them to root cause across the stack, and resolve them with operational context from across your environment. Backed by $30M Series B led by General Catalyst, along with Sierra Ventures, Bowery Capital, Epic Ventures, and Hitachi Ventures, the company is focused on moving enterprises from passive monitoring to real-time, autonomous incident resolution.
In this onsite role in San Francisco, CA, the Applied AI Engineer, Agents will combine applied research with hands-on engineering to build and ship agent systems for complex enterprise data problems. The work centers on developing agentic products, running rigorous experiments and evaluations, and translating the strongest approaches into production capabilities.
What you’ll do
- Build and ship AI agents that address complex enterprise problems and deliver meaningful customer value.
- Advance model and agent capabilities through fine-tuning, post-training, and other adaptation techniques to improve performance on enterprise tasks.
- Design experiments and evaluations, including building datasets, benchmarks, and feedback loops to track progress and diagnose failure patterns.
- Build the systems behind the agents, developing services, integrations, and infrastructure that support reliable execution at scale.
- Bring research into production by owning implementation, deployment, and ongoing improvement for promising experiments.
- Improve production performance with practical tradeoffs across quality, reliability, latency, cost, and security.
- Shape product and research direction by partnering with engineering, product, and customer-facing teams to identify valuable problems and pursue approaches that work.
What you bring
- Experience building and shipping AI systems, with substantial hands-on contributions across code, experimentation, and production engineering.
- Hands-on experience with post-training tools such as PyTorch, Hugging Face Transformers, TRL, and PEFT, including supervised fine-tuning, parameter-efficient adaptation (such as LoRA), and preference optimization or reinforcement learning.
- Practical experience with LLMs, agent systems, and model evaluation.
- Experience adapting or training models, including understanding data quality, experimental design, and generalization.
- Exposure to training and evaluation data pipelines, synthetic data generation, or human feedback systems.
- A working knowledge of agent orchestration, tool use, retrieval, or inference optimization.
- Comfort working with enterprise data platforms, distributed systems, or production AI infrastructure.
- Familiarity with distributed GPU training and efficient inference using tools such as FSDP, DeepSpeed, or vLLM, along with reproducible experimentation and evaluation.
- Strong software engineering fundamentals and proficiency in Python, TypeScript, or comparable languages.
- Ability to read research, reproduce useful results, and assess whether an approach will translate to real-world performance.
- Comfort owning ambiguous problems and moving between research exploration and product delivery.
- Clear communication, technical judgment, and a collaborative working style.
Tools and technologies
Python, TypeScript, PyTorch, Hugging Face Transformers, TRL, PEFT, LoRA, LLMs, FSDP, DeepSpeed, vLLM
Equal opportunity and accommodations
Pantomath is an Equal Opportunity Employer. Employment decisions are made without regard to legally protected characteristics and are based on qualifications, merit, and business needs. The company is committed to providing reasonable accommodations to qualified individuals during the application and interview process or throughout employment. To request an accommodation, contact the Head of People.
Pay range and FLSA status
This position is classified as Exempt under the Fair Labor Standards Act (FLSA), meaning it is not eligible for overtime compensation. Pay range: $150,000 - $230,000 USD per year.
