We are seeking a skilled AI Engineer with hands-on experience post-training large language models at scale. This is a part-time, fully remote engagement for a practitioner who has done the work, not just studied it. You'll join a focused team working toward fine-tuning and aligning a 70B-parameter model, and your prior experience at that scale (or close to it) is essential.
Eligibility requirement
Applicants must be US citizens currently residing in the United States. We are unable to consider applicants based outside the US or those without US citizenship, regardless of work authorization status.
What you'll do
- Lead and contribute to post-training workflows including supervised fine-tuning, instruction tuning, DPO, RLHF, RLAIF, and related alignment techniques
- Apply QLoRA and other efficient fine-tuning methods to models across the 7B to 70B+ parameter range
- Train models for reliable structured output generation under adversarial input conditions
- Build and operate evaluation pipelines for safety-critical model behavior, including adversarial test suites, red-team integration (e.g., Garak), and regression tracking across model versions
- Calibrate decision thresholds against tiered policy configurations, including logprob-based confidence calibration at the serving layer
- Design training approaches that preserve inference-time policy specification, so model behavior can be adjusted without retraining
- Curate and prepare training data, evaluation sets, and preference data pipelines
- Iterate on training strategy to improve task performance, calibration, and adversarial robustness
- Document approach and decisions clearly for an async-first team
What we're looking for
- US citizen currently residing in the United States. This is a firm requirement.
- Concrete, verifiable production experience post-training open-weight LLMs. Experience at 7-8B, 13-30B, or 30B+ scales is all welcome, with larger-scale work a plus.
- Experience with modern open-weight model families (Llama, Qwen, Mistral, or similar)
- Hands-on experience with QLoRA, LoRA, and efficient fine-tuning techniques for large models
- Experience with alignment methods: SFT, DPO, RLHF, RLAIF, or constitutional AI approaches
- Experience training models for reliable structured output (JSON, schema-constrained generation, function-call style outputs)
- Familiarity with modern serving stacks such as vLLM, TGI, or similar, and comfort working with logprob-level model outputs
- Deep familiarity with distributed training frameworks such as DeepSpeed, FSDP, Megatron-LM, or similar
- Proficiency in Python and comfort with multi-GPU, multi-node training infrastructure
- Ability to work independently and manage your time in a part-time, async-first environment
Strong preferences
- Direct experience training safety classifiers, content moderation models, jailbreak or prompt injection detectors, or other trust-and-safety ML systems
- Experience with adversarial evaluation frameworks (Garak, promptfoo, or similar)
- Comfort with deployment constraints typical of regulated or restricted-network environments
Bonus qualifications
- Published research or open-source contributions in LLM training, alignment, or AI safety
- Prior work at an AI lab, a foundation model team, or on a production safety classifier
- Experience designing or operating tiered policy systems where model behavior can be modulated at inference time
Work Location: Remote
Pay: $100.00 - $200.00 per hour
Work Location: Remote
