Member of Technical Staff - AI Engineer

NeoSigma is a product-driven research lab building the intelligence layer that helps close the feedback loop between your customers, products, and AI systems.

We are a small, intensely technical team of researchers and engineers who have trained frontier-scale models and widely used AI products and agents at MIT, Parallel Web, Essential AI, Apple, and Amazon.
We are backed by world-class investors and leaders from Google DeepMind, OpenAI, Decagon and others.

• →Build SOTA agentic pipelines that mine production traces for failures mining, triaging, evaluations, and optimizations
• →Design and implement large-scale data processing systems that turn production agent traces into signal for evaluation and optimization
• →Design and run optimization loops that convert evaluation signal into measurable gains in agent behavior
• →Own the cost and latency profile of our ML systems as trace volume and customer count scale
• →Push novel evaluation and post-training research into production within weeks

• Hands-on experience shipping LLM-powered or agentic systems to real production environments
• Strong instincts around evaluation design, post-training methods, and agent triaging
• Comfortable working directly with customers and translating their constraints into system decisions
• High bar for code quality, modularity, and systems that stay clean as they grow
• Ability to move fast in ambiguous territory and turn open problems into clear, executable plans
• Bachelor’s degree in Computer Science, Engineering, or a related field (or equivalent practical experience)

• Customer Obsession - We start with the customer and work backwards. We aim to earn trust through consistent delivery, thoughtful listening, and by obsessing over customers.
• Intellectual Honesty - We operate with high trust and low ego. Ideas matter more than titles, and we communicate openly and directly while assuming good intent, even in strong disagreement.
• Bias for Action - We set high standards and move quickly to meet them. We prefer building and learning with customers over debating in the abstract, and we iterate based on real feedback.
• Extreme Ownership - We take responsibility for outcomes, not just tasks. Ownership means seeing problems through to completion and ensuring solutions truly work in practice.