Forward Deployed AI Engineer

About Millennium

Millennium is a global, diversified alternative investment firm, founded in 1989. Defined by evolution, innovation and focus, Millennium’s mission is to deliver results for our investors.

Our people are empowered with both independence and support: the autonomy to pursue ideas with conviction and the backing of a global network committed to collaboration, disciplined risk management and continuous learning. With opportunities to deepen expertise and accelerate development, talent at Millennium is equipped to adapt, evolve and build lasting impact over time. Discover how transformative growth accelerates impact.

Meet the Team

Core to the health and growth of our business, Millennium’s Information Technology organization develops the flexible, scalable technology and advanced proprietary systems that support the firm’s active, multi-manager platform, including next-generation analytical and trading capabilities. Our Central AI team has a broad mandate: bring AI to every function that runs the firm. That includes business development, alpha capture, execution services, risk management, finance, operations and middle office, legal, compliance, recruiting, and beyond. If a team here has a problem AI can solve, we figure out the right approach and ship the solution.

Our Israel office is located in the Bursa area of Ramat Gan.

This role is on-site.

As a global firm, proficiency in English is required.

About The Role

We're hiring Rapid Application Developers: forward deployed AI engineers who embed with a business team, learn how the work gets done, and build the system themselves. Most of what we ship is one of two things. The first is agentic systems that execute multi-step business processes end to end. The second is supervised machine learning over years of the firm's own labeled data. This is not a chatbot role. Sometimes the right answer is a simple script, and you'll say so.

Problems arrive as "this takes my team 20 hours a week," or "we have ten years of this data and no idea what's in it." You own everything from that conversation through architecture, deployment, evaluation, and iteration, then turn what works into reusable capabilities for the whole firm.

Who thrives here: builders who run their own lives on agents they made, track model releases because the deltas change what they can ship, and treat an AI system without an eval as a demo. Equally welcome are engineers with deep classical ML and algorithm backgrounds (perception, ranking, forecasting, optimization) now applying that rigor to agentic systems. The best people on this team hold both.

What you get: frontier model access across every major provider with real budget, years of proprietary labeled data as eval and training fuel, ship cycles measured in days rather than quarters, demanding users who will actually run your systems, and no platform team between you and production.

What You'll Do

• Gentic process automation (the core of the job). Reverse-engineer manual, undocumented workflows and build agents that execute them reliably: tool integrations, orchestration, human-in-the-loop checkpoints, state management, auditability.
• Prediction and decision support. Forecasting, anomaly detection, ranking, matching, and classification over data the firm has accumulated for years, with proper validation and monitoring.
• Document intelligence. Extraction, classification, and review pipelines over contracts, filings, emails, and transcripts.
• Entity intelligence and self-improving systems. Knowledge graphs, entity resolution, enrichment pipelines, and eval/feedback loops that make deployed systems better over time.
• You might spend one quarter on an agentic workflow for execution services and the next on a forecasting model for a finance team. If you need a predictable lane, this isn't the role.
  
  

• 3+ years shipping production software, ML, or AI systems (flexible for exceptional builders; we weight what you've shipped over years served)
• Degree in a quantitative field, or a track record that makes it irrelevant
• Strong Python and engineering fundamentals; your prototypes become products others build on
• Substantial hands-on LLM/agentic experience: tool use, orchestration, structured outputs, retrieval, evals, and calibrated judgment about what agents can reliably do
• Supervised ML depth: training, validation, feature engineering, and the pitfalls (leakage, overfitting, drift)
• Comfort with messy, heterogeneous data and unfamiliar domains: you get functionally fluent in days and earn credibility with experts
• Autonomy: you can sit with a trader's team or a CFO's team, understand their world, then build without waiting for a spec