VP of Research, Machine Learning
Summary
Executive-level AI research role at an AI startup building a proactive assistant that completes real-world tasks. You define the research roadmap (memory, reasoning, planning, agent orchestration), choose between custom architectures and foundation models, and own evaluation/alignment/safety. Stack: Python, PyTorch/JAX, GPU training and inference.
Our client is an AI startup building a proactive AI assistant that helps users complete real-world tasks through intelligent, goal-driven workflows.
As the Head of AI Research, you will define the long-term intelligence strategy behind the product. You'll lead the research direction for how AI understands context, reasons, plans, evaluates, and continuously improves in a product used frequently by real users.
Responsibilities
- Define and evolve the research roadmap for the platform's core intelligence, including context representation, memory, reasoning, planning, and agent orchestration.
- Determine when to develop new model architectures versus adapting or leveraging frontier open-source and commercial foundation models.
- Design evaluation frameworks that measure real-world usefulness, robustness, safety, and long-term system behavior—not just benchmark performance.
- Own the strategy for AI alignment, safety, and guardrails as core product capabilities.
- Lead the exploration and application of advanced AI techniques, including: Retrieval-augmented training (RAT), Mixture-of-Experts (MoE), Model distillation, Multi-agent orchestration and Multimodal AI systems.
- Partner closely with product, machine learning, and engineering teams to shape the product's intelligence roadmap and long-term technical vision.
- Set a high standard for research quality, technical judgment, experimentation, and engineering excellence across the organization.
Requirements
- Extensive experience building, deploying, or evolving machine learning systems in production.
- Deep understanding of modern machine learning, large language models, reasoning systems, and AI architectures.
- Strong technical judgment around model behavior, failure modes, and long-term architectural trade-offs.
- Proven ability to translate research into practical, production-ready AI systems.
- Comfortable making high-impact technical decisions in fast-moving, ambiguous environments.
- Strong focus on evaluation, correctness, safety, and long-term system performance.
- A builder's mindset with a high degree of ownership—you care about delivering systems that work reliably in the real world, not just advancing research.
- Operates like a founder: proactive, accountable, and driven to solve difficult technical problems from first principles.
- Technologies and tools: Python, PyTorch / JAX, GPU-accelerated training and inference systems