Impress employers and recruiters.
Choose from hundreds of resume examples.

Target Resume Now

About the Role

We are hiring a software engineer to build and scale the distributed systems powering Tinker, our post-training platform. You will work on the infrastructure for scheduling jobs across large GPU clusters, maintaining state consistency under failure, and providing a reliable path from idea to a customized model. This role is central to Tinker's reliability and scale, involving system design, operation, and incident response with significant ownership and minimal oversight.

Responsibilities

Design, build, and operate the distributed systems underlying Tinker, including job scheduling, resource orchestration, checkpointing, and fault recovery across large multi-node GPU clusters.
Own the reliability, latency, and throughput of live production systems serving concurrent training and sampling workloads for external users.
Build for graceful degradation and fast recovery: handle node failures, preemptions, and network partitions without losing user progress or data.
Improve observability across the platform — metrics, tracing, and alerting that make it possible to detect and diagnose issues before or as they affect users.
Lead incident response for the systems you own, drive root-cause analysis, and turn findings into lasting fixes.
Partner with research and ML infrastructure teams to understand new training and sampling workloads and to evolve the platform's architecture to support them at scale.
Make build-vs-buy and architectural tradeoff calls for core infrastructure components, and mentor other engineers on distributed systems practices.

Requirements

7+ years of experience building, running, and scaling distributed systems in production, with direct on-call ownership of live services.
Deep understanding of distributed systems fundamentals: consensus, consistency models, failure detection, scheduling, and state management under partial failure.
Track record of designing systems that operate reliably at scale, including diagnosing and resolving production incidents under time pressure.
Strong systems-level coding ability in a language such as Python, Go, Rust, or C++, and comfort working close to the infrastructure layer.
Experience operating large-scale GPU or accelerator clusters, including job orchestration, resource scheduling, or ML training infrastructure.
Familiarity with the training or inference stack for large language models, and an understanding of the systems demands unique to fine-tuning or post-training workloads.
Experience with checkpointing, distributed storage, or high-throughput networking (e.g., RDMA, NCCL) in performance-critical settings.
Experience building developer-facing infrastructure or APIs where reliability and usability are both first-class concerns.
A track record of driving platform-wide architectural decisions and comfort operating with significant autonomy in a fast-changing environment.

Skills

Python
Go
Rust
C++
Distributed Systems
Job Scheduling
Resource Orchestration
Checkpointing
Fault Recovery
Reliability
Latency
Throughput
Observability
Incident Response
Root-Cause Analysis
ML Infrastructure
GPU Clusters
Large Language Models
Fine-tuning
Distributed Storage
High-throughput Networking
RDMA
NCCL

Location

San Francisco, California

Work Type

Full-time

Experience Level

7+ years

Salary/Compensations

$350,000 - $475,000 USD

Benefits

Generous health, dental, and vision benefits
Unlimited PTO
Paid parental leave
Relocation support

About the Company

The mission of Thinking Machines is to build AI that extends human will and judgment.
We are training frontier models with Inkling, developing Tinker to let people make models their own, and crafting interfaces that broaden human-AI communication.
We believe the future worth building is human, and we're hiring people who want to build it.
Tinker is our fine-tuning API that empowers researchers and developers to customize frontier AI to their needs — opening access to capabilities that have previously been concentrated in a handful of labs.
We manage the infrastructure while allowing Tinkerers full flexibility in training open weights models with their own data, algorithms, and for their own needs.
Tinker is rapidly adding new customers, features, and novel use-cases, and the systems underneath it run as a live, always-on production platform.