About the job
Job Description
Embrace the future of competitive advantage with Eragon, where we create bespoke AI systems that are meticulously tailored to understand your unique business landscape.
At Eragon, we focus on developing AI models that leverage proprietary data, deployed directly within customer environments and continuously refined through real-world interactions. Our models not only respond but evolve, improving with each user engagement.
We utilize a cutting-edge reinforcement learning framework known as RLQF (Reinforcement Learning from Query Feedback) that transforms user interactions into valuable training signals, establishing a cycle of ongoing enhancement that surpasses traditional fine-tuning methods.
The Role
As an Applied Research Engineer, you will be responsible for designing, training, and deploying advanced models that drive real business operations.
This position is not about theoretical research; you will engage directly with customer data, constraints, and feedback, crafting solutions that excel in production settings. You will manage the entire lifecycle of the project, from defining the problem and designing data structures to training, evaluating, and iterating based on live performance.
What You’ll Do
- Train and adapt models: Fine-tune and post-train models on customer-specific data utilizing RLQF among other techniques.
- Close the loop: Convert real user interactions, corrections, and workflows into actionable training signals.
- Own end-to-end systems: Oversee the process from data ingestion and curation through to training, evaluation, and deployment.
- Evaluate in production: Create evaluation frameworks that accurately reflect real-world performance, rather than relying solely on benchmarks.
- Work with customers: Collaborate closely with users to comprehend their workflows and translate these into model functionalities.
- Ship and iterate: Focus on the continuous improvement of models based on live feedback and measurable outcomes.
What We’re Looking For
- Extensive hands-on experience in training, fine-tuning, or post-training machine learning models.
- Proficiency in handling messy, real-world data as opposed to only clean benchmarks.
- Familiarity with reinforcement learning techniques, feedback-driven training such as RLHF or RLAIF, and evaluation systems.
- Adeptness at quickly transitioning from problem identification to data management, model development, and iterative improvement.
- Strong engineering instincts with a comfort level in managing systems end-to-end.
- A proactive approach to shipping and enhancing systems, rather than solely focusing on research.
