Hireclout
Member of Technical Staff – Model Training
Hireclout, Palo Alto, California, United States, 94306
Job Title: Member of Technical Staff – Model Training
Role Overview Join a fast-moving AI company focused on enterprise-grade conversational intelligence. This organization is a mission-driven public benefit corporation that equips businesses with customizable language models, proprietary data pipelines, and intelligent tuning systems—allowing virtual assistants to become smarter, more accurate, and brand-aligned over time.
This role sits at the intersection of ML research and production engineering. As a Model Training Engineer, you’ll help turn general-purpose LLMs into finely tuned, high-performing assistants using cutting-edge post-training and fine-tuning techniques. You'll have access to massive GPU clusters, real-world feedback loops, and the autonomy to experiment, iterate, and deploy improvements rapidly.
Key Responsibilities
Develop and maintain scalable post-training workflows including dataset curation, evaluation, hyperparameter tuning, and rollout
Experiment with and deploy advanced alignment methods such as RLHF, DPO, GRPO, and RLAIF
Build training automation tools, dashboards, and pipeline components to improve reproducibility and traceability
Define key training metrics, run A/B tests, and quickly iterate to hit performance goals
Collaborate cross-functionally with inference, safety, and product teams to integrate model improvements into user-facing systems
Education & Qualifications Hands-on experience training large transformer models on distributed GPU systems (multi-GPU, multi-node)
Strong proficiency with Python and PyTorch, including ecosystem tools like Torchtune, FSDP, and DeepSpeed
Practical understanding of reinforcement learning techniques (RLHF, DPO, GRPO, RLAIF)
Effective communicator across both technical and non-technical stakeholders
Proven ability to build reproducible and automated training infrastructure
Preferred Experience Experience with multimodal (vision-language, audio-text) or voice models
Familiarity with cross-modal data preparation and model alignment
Contributions to open-source ML tooling
Why Us High-impact mission – Shape the future of enterprise AI by building assistants that reflect each brand’s voice authentically
Massive compute resources – Access thousands of NVIDIA and Intel Gaudi GPUs for rapid iteration and experimentation
Growth & autonomy – Competitive compensation ($200K–$350K base), meaningful equity, and ownership of critical projects
Open-source culture – Actively contribute to projects like Torchtune, PyTorch, and vLLM; every engineer is encouraged to give back
Applicants must be currently authorized to work in the United States on a full-time basis now and in the future. This position does not offer sponsorship.
#J-18808-Ljbffr
Experiment with and deploy advanced alignment methods such as RLHF, DPO, GRPO, and RLAIF
Build training automation tools, dashboards, and pipeline components to improve reproducibility and traceability
Define key training metrics, run A/B tests, and quickly iterate to hit performance goals
Collaborate cross-functionally with inference, safety, and product teams to integrate model improvements into user-facing systems
Education & Qualifications Hands-on experience training large transformer models on distributed GPU systems (multi-GPU, multi-node)
Strong proficiency with Python and PyTorch, including ecosystem tools like Torchtune, FSDP, and DeepSpeed
Practical understanding of reinforcement learning techniques (RLHF, DPO, GRPO, RLAIF)
Effective communicator across both technical and non-technical stakeholders
Proven ability to build reproducible and automated training infrastructure
Preferred Experience Experience with multimodal (vision-language, audio-text) or voice models
Familiarity with cross-modal data preparation and model alignment
Contributions to open-source ML tooling
Why Us High-impact mission – Shape the future of enterprise AI by building assistants that reflect each brand’s voice authentically
Massive compute resources – Access thousands of NVIDIA and Intel Gaudi GPUs for rapid iteration and experimentation
Growth & autonomy – Competitive compensation ($200K–$350K base), meaningful equity, and ownership of critical projects
Open-source culture – Actively contribute to projects like Torchtune, PyTorch, and vLLM; every engineer is encouraged to give back
Applicants must be currently authorized to work in the United States on a full-time basis now and in the future. This position does not offer sponsorship.
#J-18808-Ljbffr