Senior Applied AI Manager

Описание вакансии

Oumi · Hybrid (Seattle / SF / NY) · Full-time

About Oumi

Why we exist:

Oumi is on a mission to make frontier AI truly open for all. We believe that AI will have a transformative impact on humanity. As such, AI should be developed openly and collectively. It should be made universally accessible.

What we do:

Oumi provides an end-to-end, AI-native platform to build custom AI models in hours, not months –automating the loop of evaluation, data synthesis, training, and repeat. Oumi also develops an open research stack and models in collaboration with academic collaborators and the open community.

The Role

We're looking for a Senior Applied AI Manager to own the strategy and execution for AI science at Oumi. This is a senior AI science leadership role in the company: you'll set the applied science agenda, build and lead the team, and be accountable for the science quality of every feature that ships on our platform.

Your scope spans the full model development lifecycle—data strategy, pre-training and post-training methodology, evaluation science, and production deployment—as well as the agentic systems that automate and improve each stage. You'll work closely with the CEO and product leadership to translate Oumi's company strategy into a concrete AI science roadmap, then execute against it with a growing team of ML engineers and applied researchers.

This role blends research and product shipping. You'll stay very close to the academic research, but also industry trends. You will leverage AI science, drive experimentation, and translate breakthroughs into production systems that Oumi and our customers use every day.

What You'll Do

• AI Science Strategy & Roadmap: Define and drive the research and engineering roadmap for AI science at Oumi. Translate company objectives into concrete milestones for model quality, capability, and efficiency—and make the hard prioritization calls when resources are scarce.
• Team Building: Recruit, manage, and develop a high-performing team of ML engineers and applied researchers. Set a high bar for talent, create an environment of rigorous experimentation, and coach people toward increasing scope and independence.
• Training Science: Lead experimentation across the full training stack—pre-training, supervised fine-tuning, alignment (RLHF, DPO, GRPO), distillation, curriculum learning, and data mixing—to systematically improve model quality with each generation.
• Data Strategy: Own the data side of model development. Build intelligent pipelines for quality scoring, filtering, deduplication, and synthetic data generation. Develop a data-scientific understanding of what data actually moves the needle and use it to guide investment.
• Evaluation & Feedback Loops: Design evaluation frameworks that go beyond static benchmarks. Build automated feedback loops where evaluation signals inform data selection, training decisions, and agent behavior—creating a flywheel of continuous improvement.
• Agentic Workflows: Research and develop agent-based systems that orchestrate the model training lifecycle—from automated hyperparameter optimization to self-improving data curation—so training runs get smarter over time with less manual intervention.
• Production & Deployment: Partner with infrastructure and product teams to ensure AI science features ship reliably, perform at quality.
• Open Source & Community: Publish findings, contribute to open-source tooling, and collaborate with external researchers and academic partners. Represent Oumi's AI science work in the broader research community.

What You'll Bring

• Experience: 5+ years of professional experience in ML research, ML engineering, or a closely related field. Demonstrated track record of turning research into production systems. PhD in AI or equivalent industry experience.
• Management: 1+ years of experience managing engineers or applied researchers. You've hired, coached, and retained strong technical talent.
• ML Depth: Expertise across the model training lifecycle—pre-training, fine-tuning (SFT, RLHF, DPO), evaluation, and deployment. Hands-on experience training or substantially improving LLMs or VLMs.
• Research Mindset: You design rigorous experiments, interpret results critically, and stay current with the literature. You know when to apply an existing technique and when to invent something new.
• Agentic Systems: Experience building or working with LLM-powered automation, tool-use patterns, or multi-agent architectures. You think naturally about how to decompose complex tasks into agent-friendly steps.

Nice to Have

• Ph.D. in Computer Science, Machine Learning, or a related field.
• Publications in ML/AI venues (NeurIPS, ICML, ICLR, ACL, etc.).
• Experience with data-centric ML approaches—data quality estimation, curriculum learning, or synthetic data generation.
• Contributions to open-source ML frameworks or tooling.
• Familiarity with ML infrastructure (Kubernetes, GPU clusters, orchestration frameworks).
• Prior experience at an early-stage or high-growth startup where you wore multiple hats across research, engineering, and strategy.