Remote ML Engineer — End-to-End Fine-Tuning

Run BIOS INC.; San Francisco (CA); Hybrid; USD 180,000 - 240,000; Full time

Job summary

Run BiOS is seeking an Applied Machine Learning Engineer to fine-tune open models on dedicated GPUs, billed per second, and to deploy them via a single API. You will own end-to-end fine-tuning engagements, build internal tooling for validation and orchestration, and help customers translate problems into platform capabilities.

Strong Python/PyTorch skills and clear communication are essential. You will work with datasets before modeling, debug challenging issues like loss spikes and tokenizer

Qualifications

• 5+ years applying machine learning in production environments.
• Experience fine-tuning LLMs or vision-language models — configuring runs, reading curves, not only calling APIs.
• Strong Python and PyTorch, and comfortable reasoning about GPU memory when things get tight.
• Data instincts: you look at the dataset before you touch the model.
• Clear communication — you can explain a failed run to a customer without hand-waving.

Responsibilities

• Own end-to-end fine-tuning engagements: dataset inspection, run configuration, evaluation, and handoff.
• Build tooling for dataset validation, training orchestration, and model evaluation.
• Debug loss spikes, adapter pathologies, tokenization edge cases, and VRAM limits.
• Summarize customer patterns to product and research as prioritized proposals.
• Assist customers in serving fine-tuned weights via our inference stack.

Skills

Python

PyTorch

Data instincts

LLM fine-tuning

Memory management

Tools

Training orchestration tooling

Model evaluation tooling

Job description

Run BiOS is seeking an Applied Machine Learning Engineer to fine-tune open models on dedicated GPUs, billed per second, and to deploy them via a single API. You will own end-to-end fine-tuning engagements, build internal tooling for validation and orchestration, and help customers translate problems into platform capabilities.

Strong Python/PyTorch skills and clear communication are essential. You will work with datasets before modeling, debug challenging issues like loss spikes and tokenizer