Full-Time
ML Research Engineer
Generative AI, Large Language Models
Posted on 7/25/2025
Oumi
11-50 employees
Enterprise open AI platform with customization
No salary listed
Palo Alto, CA, USA + 2 more
More locations: Seattle, WA, USA | New York, NY, USA
Hybrid
Hybrid work environment with offices in Seattle, WA, Palo Alto, CA, and New York, NY.
 AI & Machine Learning (3) |
|---|
| , , |
- Strong experience in machine learning, deep learning, or NLP—especially in generative AI or LLMs.
- Solid programming skills in Python, and experience with ML frameworks like PyTorch.
- Experience designing or maintaining ML infrastructure at scale (e.g., cloud-based training, distributed systems).
- Comfort working in highly collaborative environments with research and engineering teams.
- Design and build systems to support training, fine-tuning, and evaluating large language models.
- Partner with researchers to define experiments, write reusable code, run benchmarks, and interpret results.
- Work on LLM alignment and tuning using techniques like reinforcement learning (RLHF), supervised fine-tuning, and prompt optimization.
- Develop scalable ML pipelines for distributed training (e.g., across multi-GPU and multi-node environments).
- Contribute to open-source tooling and models to support transparency and community collaboration.
- Optimize performance across the ML stack—from data loading to deployment.
- Experience with academic publications, open-source contributions, or LLM alignment work.
Oumi.ai provides enterprise-grade, open AI platforms designed to remove vendor lock-in and reduce fragmented dependencies. The company works with enterprises and communities to develop custom AI solutions, partnering with businesses to advance AI research and create humane, widely accessible AI capabilities. Revenue comes from enterprise solutions and ongoing support services. The platform enables building and deploying foundation models through flexible integration options, configurable training environments, and evaluation benchmarks, allowing organizations to tailor AI models to their needs while ensuring safety and governance. What sets Oumi.ai apart is its emphasis on openness, collective development, and strong customer ownership, offering integrated tools and services that support both research and production at scale. The company aims to make AI broadly accessible by fostering open collaboration and reliable, enterprise-ready AI capabilities that teams can trust and extend.
Company Size
11-50
Company Stage
Seed
Total Funding
$10M
Headquarters
Seattle, Washington
Founded
2024
Simplify's Take
What believers are saying
- Raised $10M seed from Venrock and Obvious Ventures fueling 20-employee growth.
- Aurasell’s 8B model beat Sonnet 4.5 by 8% coverage using Oumi platform.
- 9,000 GitHub stars and adoption by Stanford, MIT, Berkeley drive expansion.
What critics are saying
- OpenAI o1-mini-custom automates fine-tuning at 70% of Oumi’s compute cost.
- Google DeepMind’s ModelForge free TPUs adopted by Stanford, MIT labs.
- HallOumi-8B fails 25% adversarial tests per Berkeley benchmark March 2026.
What makes Oumi unique
- HallOumi-8B outperforms OpenAI o1 and Claude Sonnet 3.5 at 8B parameters.
- Automates end-to-end custom model development from natural language prompts.
- Enables per-sentence hallucination detection with citations and explanations.
Help us improve and share your feedback! Did you find this helpful?
Benefits
Health Insurance
Dental Insurance
Vision Insurance
Paid Vacation
Company Equity
Company News
Google's ai-generated answers often inaccurate, New York Times investigation reveals. The reliability of summaries produced by artificial intelligence has once again become a major concern. This issue is all the more crucial as it pertains to information access for a vast proportion of the global population. A comprehensive study, conducted by The New York Times in collaboration with the startup Oumi, provides essential clarification on this matter. This investigation utilized a corpus of 4,000 questions to precisely evaluate the performance of AI...
Oumi aims to simplify and automate custom AI model development. Oumi PBC, a startup promoting an open artificial intelligence platform built in concert with researchers from some of the world's most prominent universities, today introduced a platform designed to automate the development of custom AI models, positioning it as a way for enterprises to move beyond reliance on large, general-purpose offerings. The announcement builds on what the company says is the momentum its earlier open-source project has built, including nearly 9,000 positive ratings on GitHub and adoption by dozens of research institutions. The new platform targets a broader audience, particularly enterprise teams that lack the time or expertise to build models from scratch. The company's pitch centers on automating what has traditionally been a complex, multistep process. Enterprises increasingly want to move away from large, closed models toward "small language models," specialized alternatives tailored to their specific needs, said Chief Executive Manos Koukoumidis (pictured). "We've seen a wave of enterprises that started with large, off-the-shelf models but now want to move to open specialized SLMs," which offer benefits such as greater relevance to specific projects, lower cost and lower latency, Koukoumidis said. That transition has been slowed by the difficulty of building custom models, a process that can take months, he said. End-to-end automation. Oumi's platform is designed to address that barrier by automating the end-to-end workflow, including data generation, evaluation, training and iteration, which are "all the steps that an AI engineer takes to develop a custom model," Koukoumidis said. The company claims the system can build custom AI models up to 100 times faster than conventional processes, reducing weeks or months of human effort to hours or minutes. Oumi said it can also reduce the technical barriers to model generation by allowing users to define a task in natural language. "All you need to do is say, 'I want to build a model,' and then you kick it off," Koukoumidis said. The platform then automates key tasks such as defining evaluation metrics, generating synthetic data and refining models based on performance gaps. "It analyzes the results for you, automatically synthesizes data, fine-tunes the model and keeps iterating," he said. "It automates all the steps of an engineer." The system supports a wide range of open models. Pricing is based on usage, including compute and tokens for training and inference. Oumi said the cost is offset by reduced engineering effort. "The amount of human effort that it saves is much higher than the bill you're going to get to train," Koukoumidis said. Oumi, which operates as a public benefit corporation, frames the platform as part of a broader effort to decentralize AI development. Koukoumidis said the company's goal is to "democratize the use of AI and put the future of AI in the hands of enterprises." Photo: oumi. A message from John Furrier, co-founder of SiliconANGLE: Support its mission to keep content open and free by engaging with theCUBE community. Join theCUBE's Alumni Trust Network, where technology leaders connect, share intelligence and create opportunities. * 15M+ viewers of theCUBE videos, powering conversations across AI, cloud, cybersecurity and more * 11.4k+ theCUBE alumni - Connect with more than 11,400 tech and business leaders shaping the future through a unique trusted-based network. About SiliconANGLE Media SiliconANGLE Media is a recognized leader in digital media innovation, uniting breakthrough technology, strategic insights and real-time audience engagement. As the parent company of SiliconANGLE, theCUBE Network, theCUBE Research, CUBE365, theCUBE AI and theCUBE SuperStudios - with flagship locations in Silicon Valley and the New York Stock Exchange - SiliconANGLE Media operates at the intersection of media, technology and AI. Founded by tech visionaries John Furrier and Dave Vellante, SiliconANGLE Media has built a dynamic ecosystem of industry-leading digital media brands that reach 15+ million elite tech professionals. Its new proprietary theCUBE AI Video Cloud is breaking ground in audience interaction, leveraging theCUBEai.com neural network to help technology companies make data-driven decisions and stay at the forefront of industry conversations.
Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More. In the race to deploy enterprise AI, one obstacle consistently blocks the path: hallucinations. These fabricated responses from AI systems have caused everything from legal sanctions for attorneys to companies being forced to honor fictitious policies. Organizations have tried different approaches to solving the hallucination challenge, including fine-tuning with better data, retrieval augmented generation (RAG), and guardrails. Open-source development firm Oumi is now offering a new approach, albeit with a somewhat ‘cheesy’ name.The company’s name is an acronym for Open Universal Machine Intelligence (Oumi). It is led by ex-Apple and Google engineers on a mission to build an unconditionally open-source AI platform.On April 2, the company released HallOumi, an open-source claim verification model designed to solve the accuracy problem through a novel approach to hallucination detection
Today, Oumi PBC is excited to introduce HallOumi-8B and HallOumi-8B-Classifier, a family of open-source claim verification (hallucination detection) models, outperforming DeepSeek R1, OpenAI o1, Google Gemini 1.5 Pro, Llama 3.1 405B, and Claude Sonnet 3.5 at only 8 billion parameters!
Oumi co-founders Manos Koukoumidis, left, and Oussama Elachqar. (Oumi Photo) A new startup out of Seattle wants to open up the "black box" of foundational