5 hours ago
Base Salary
$230k - $325k/yr
Responsibilities
- Create high-quality training datasets from prompts, responses, citations, and query fanouts.
- Build and iterate on supervised fine-tuning, preference optimization, and reinforcement learning pipelines.
- Design reward signals using reward models, LLM judges, and verifiable rewards.
- Train models for tool use and long-horizon agentic workflows with AI Marketer and analytics teams.
- Define model evaluations with the Evals team and assess every training run against them.
- Prototype post-training methods on compute and data infrastructure and move successful approaches into production.
- Deploy trained models, own their production performance, and use regressions and customer behavior to guide subsequent training cycles.
Requirements
- Exceptional software engineering skills in Python and PyTorch, with JAX as a plus.
- Hands-on post-training experience with SFT, RLHF, DPO, GRPO, or similar reinforcement learning methods.
- Experience with distributed training systems such as FSDP, DeepSpeed, Ray, Kubernetes, or Slurm.
- Strong evaluation judgment and the ability to identify misleading metrics.
- Demonstrated ownership of work from early experimentation through production model deployment.
- Ability to work effectively in a fast-paced startup with rapid iteration cycles.
- Preferred experience with agentic or tool-use training, synthetic data, reward model design, retrieval, or search-grounded models.
- Preferred publications at venues including NeurIPS, ICML, ICLR, MLSys, ACL, or EMNLP.
Benefits
- On-site work in the New York City or San Francisco office.
- Base salary range of $230,000 to $325,000, plus equity and a full range of benefits and perks.
Tech Stack
Categories
About Profound
Profound builds a SaaS marketing intelligence platform that helps enterprise brands measure and improve how they appear in AI-driven discovery across tools like ChatGPT and Perplexity. The product delivers real-time visibility analytics, competitive intelligence, and optimization insights for search-like AI channels. The company is privately held, headquartered in New York, and is Series C–funded.