Senior Machine Learning Engineer

Capital Group’s AI Insights team is hiring a Senior Machine Learning Engineer to build an insight layer over investment data using multi-agent and generative AI, backed by rigorous evaluation and applied delivery.

Responsibilities

• Clarify an underspecified request into a solvable problem, including what is being asked, what counts as an answer, and what evidence resolves it
• Extract signal from messy, incomplete datasets and distinguish real results from leakage, lucky splits, or self-flattering metrics
• Define evaluations for Generative AI performance, including eval sets, success criteria, LLM-as-judge approaches, failure modes, and alignment between measured numbers and intended meaning
• Run experiments that settle the team’s open questions and document them so decisions are reproducible, including criteria committed before results are known
• Design and build agent systems that generate insight, including task decomposition, orchestration choices, and determining where humans belong in the loop
• Identify when a single model call or a deterministic step is the most honest solution
• Build end-to-end prototypes, using AI coding tools to move quickly while keeping outputs clean and working
• Take projects from early concepts into tools people actually use, starting with short designs shaped with the team
• Raise team craft through design and code review, plus mentoring on experimental design and evaluation rigor

Requirements

• Research depth and scientific rigor: extracting real signal from ambiguous data; designing evaluation approaches; skepticism when results look unusually strong
• Abstraction and problem framing: identifying core constraints in unfamiliar problems without handholding; pursuing reusable structures over one-off solutions
• First-principles problem solving: starting from the problem and constraints rather than a preferred tool; choosing the simplest working approach
• Applied ML and Generative AI in production: delivering end-to-end systems from data understanding through evaluation to something used by others
• AI acumen: learning new tools because you want to understand how they work; working with AI coding assistants day to day; explaining what you built, where the assistants helped, and where you took over
• Communication and collaboration: clearly describing trade-offs to non-technical partners; able to state “I don’t know” without discomfort; fairly representing the other side in disagreements; improving teammates through the way you work; supporting directions you didn’t personally choose
• Ownership: driving ambiguous work to results independently
• Builder judgment: 7+ years of professional experience while staying hands-on; turning ideas into working prototypes personally; reading code with taste; using AI coding tools to produce clean results rather than accepting defaults; focusing on building enough to make research real

Preferred

• Designing and evaluating multi-agent or tool-using systems, including a clear understanding of where they fail
• Building evaluation infrastructure: eval sets, offline and online measurement, regression and drift detection
• Finance or investment management experience, or a demonstrated ability to become fluent in a new domain quickly

Experience

• Minimum experience: 7 years
• Weighting: research depth and scientific rigor; abstraction and problem framing; first-principles problem solving; applied ML and generative AI experience in production; AI acumen; communication, collaboration, and maturity; ownership; builder judgment

Compensation

• Salary range: USD 201,683 to 322,693 per year
• Location: Los Angeles, CA (hybrid)

Benefits

• Generous time-away and health benefits from day one, with flexible work options
• 2-for-1 matching gifts for charitable contributions and the opportunity to secure annual grants
• On-demand professional development resources
• Competitive salary, bonuses and benefits
• Company-funded retirement contribution factoring in salary and variable pay, including bonuses
• Individual annual performance bonus
• Capital’s annual profitability bonus
• Retirement plan where Capital contributes 15% of eligible earnings

How We Work and What We Value

• No keeping score
• Disagreement stays focused on the work; influence without needing to be the loudest
• Significant time spent in working sessions including brainstorming, design review, and pair programming
• Problems are genuinely ambiguous and not all attempts succeed
• Rigor: try to break your own results before others do
• Ownership: you act as a driver, not a passenger
• Humility: a better argument can change your mind
• Pragmatism: know when a rough answer is enough versus when it must be airtight