ML Engineer — Tilde Research

Location:

San Francisco, CA (Onsite)

Compensation:

$150,000 – $350,000 base \+ 0\.5% – 1% equity

About Tilde Research

Tilde Research is a frontier AI research lab focused on mechanistic understanding of model architectures and optimization. They work on reasoning, pretraining, scaling laws, and architecture-level experimentation. Backed with strong funding and compute resources, Tilde operates with a small, highly selective team at the intersection of research and engineering.

About The Role

This is a research engineering role focused on implementing and scaling model development work at the frontier of AI. You will translate research ideas into working systems, own end-to-end training pipelines, and operate across transformer training, distributed systems, and evaluation. The role sits between research and engineering — you're expected to read papers, implement architectures, debug large-scale training issues, and reason about model behavior and optimization at a deep level.

What You'll Own

• Implement research ideas in PyTorch or JAX and translate them into production-level systems
• Run transformer training across distributed environments
• Build and operate end-to-end pipelines covering data, training, and evaluation
• Identify and resolve divergence, instability, and throughput issues at scale
• Optimize training performance and system efficiency
• Work closely with researchers to evaluate architectures and learning dynamics
• Contribute to experimentation across model behavior and optimization strategies

Requirements

• Strong experience with PyTorch or JAX
• Hands-on experience with transformer training, distributed systems, and evaluation frameworks
• Ability to debug large-scale training issues including instability and performance bottlenecks
• Experience implementing research ideas and custom architectures
• Strong understanding of model behavior, optimization, and learning dynamics

Nice to Have

• Experience with Megatron-LM, DeepSpeed, xformers, or similar frameworks
• Background at frontier AI labs or top research engineering teams
• Open source contributions to ML frameworks
• Strong mathematical foundation in optimization or learning theory

This Role Is NOT For

• MLOps or infrastructure-focused profiles without model training experience
• Applied ML roles not involved in pretraining or model development
• Backgrounds focused on deployment rather than training systems

Logistics

• Role is fully onsite in San Francisco — please only apply if you can commit to this
• Multiple headcount \| Completely open pipeline — strong opportunity to be first in

Shortlisted candidates will be contacted by

David Joseph \& Co.

, the recruiting partner managing this search on behalf of Tilde Research.