Research Engineer, Post-Training Inference
Together AI3 months ago
Base Salary
$200k - $290k/yr
Responsibilities
- Design and build systems for customizing open-source models.
- Integrate Model Shaping and Inference platforms to connect post-training with production serving.
- Add inference-engine features and optimize reinforcement-learning workloads for large-scale post-training experiments.
- Maintain a stable, robust platform through on-call participation and 24/7 availability support.
- Collaborate with product, research, and engineering teams to keep APIs reliable, performant, and integrated with company infrastructure.
Requirements
- At least 2 years of experience building and deploying machine-learning services in production.
- Hands-on experience with modern inference engines such as SGLang, vLLM, and TensorRT-LLM.
- Familiarity with current fine-tuning methods for LLMs and other AI models.
- Strong software engineering background in Python or Go.
- Experience with low-precision FP4/FP8 serving, Multi-LoRA, or models distributed across multiple GPU nodes is preferred.
- Experience optimizing reinforcement-learning workloads is preferred.
- Experience developing CUDA, Triton, or CuTE DSL inference kernels is preferred.
- Experience developing large-scale, high-load production systems is preferred.
- Experience maintaining or contributing to open-source ML projects is preferred.
- Experience managing machine-learning workloads on Kubernetes clusters is preferred.
Benefits
- Competitive compensation, startup equity, health insurance, and other benefits.
- Full-time position; US base salary range is $200,000-$290,000.
Tech Stack
Categories
About Together AI
Together AI builds an AI-native cloud platform for developers, offering high-performance inference, fine-tuning/model shaping, and large-scale pre-training on on-demand GPU clusters with APIs and managed services. It emphasizes open-source models that teams can run and adapt, and also provides infrastructure for decentralized and scalable workloads. Founded in 2022 and headquartered in San Francisco, it is privately held and reports notable customers including Cursor, ElevenLabs, Salesforce, and Zoom.