The Role
The AI/ML Specialist will enhance ML algorithms for fraud and risk, engage with financial clients, and drive technical success with real-world data testing.
Summary Generated by Built In
Join the Kumo Team
Why This Role (and Why Now)
What You’ll Do
Minimum Qualifications
Preferred Qualifications (Bring Strength in at Least One Area)
Working Model
Success Looks Like (First 3–6 Months)
Why Join Kumo?
Compensation
Kumo is building the next generation of AI for structured data. With our Relational Foundation Model, we help some of the world’s largest companies transform their data into predictions, decisions, and end-to-end automated systems.
Kumo is already used by many of the largest consumer banking and financial institutions in the world. Our culture is collaborative, fast-moving, and deeply user-obsessed. We value people who take initiative, learn quickly, communicate clearly, and enjoy solving hard technical + people challenges.
Fraud and financial prediction are no longer “flat table” problems. The highest-value signals live in how entities connect and evolve over time: accounts ↔ devices ↔ merchants ↔ transactions, and the networks behind them.
Graph Transformers are enabling a generational leap in model quality on these highly connected datasets. The opportunity now is to bring that leap to the most sophisticated data science + ML teams in finance—teams that will pressure-test everything (leakage, ablations, calibration, robustness, drift) and only adopt what stands up in the real world.
We’re hiring an Applied ML Specialist who can do both:
- Push the frontier of our core ML algorithms for financial use cases, and
- Work directly with customers to make those capabilities real, trusted, and deployed—then feed those learnings back into the product.
This is a unique opportunity for someone who is:
- Energized by high-stakes predictive problems (fraud, risk, forecasting) where the “last mile” matters.
- Highly technical, with a research mindset and strong engineering instincts.
- Excited to be customer-facing and own outcomes—because in a startup, the best product ideas come from the field.
- Motivated by leverage: the things you build become platform capabilities used across many deployments.
Support and eventually own technical success for strategic financial services customers adopting Kumo—and convert what you learn into improvements to the core platform.
You will:
- Own a domain (fraud/AML, credit risk, or forecasting): define “what good looks like,” build the evaluation plan, run the experiments, and drive adoption.
- Work hands-on with large-scale relational datasets and customer pipelines, with a focus on connected + temporal modeling.
- Design and execute rigorous model validation: leakage-proof evaluation, calibration, robustness to drift/adversaries, and practical interpretability for real teams.
- Translate ambiguous customer needs into concrete modeling workflows and rollout plans.
- Partner closely with Kumo engineering and research to ship platform improvements informed by real customer constraints.
- Act as a technical leader and trusted advisor, understanding that deploying ML is as much a people and business challenge as it is a technical one.
- Deliver demos, workshops, best practices—and help drive pilots → production → expansions (including technical diligence during deal cycles).
Example use cases include:
- Fraud detection (rings, mule networks, ATO/CNP, abuse patterns)
- Credit scoring, risk modeling, and underwriting
- Relational forecasting across entities and time
- Financial customer analytics (propensity, retention, growth, risk-aware marketing)
- Bachelor’s, Master’s or PhD in a STEM field (CS, EE, Math, Physics, Stats, etc.) or equivalent practical experience.
- Strong fundamentals in machine learning, statistics, and data science.
- Proven ability to improve ML systems end-to-end: data → modeling → evaluation → production constraints (not just notebooks).
- Solid engineering skills: proficient developing safe and correct code with the latest coding agents.
- Strong communication skills; comfortable navigating technical + non-technical audiences.
- Motivated, self-driven, excited to learn fast, and comfortable in a high-velocity startup environment.
Deep expertise in one or more of:
- Fraud / AML / networked abuse detection (adversaries, class imbalance, delayed labels, investigations)
- Credit risk, scoring, underwriting, or lending analytics (calibration, stability, governance constraints)
- Forecasting at scale (temporal correctness, leakage control, regime shifts, multi-entity forecasting)
- Graph + temporal modeling experience (GNNs, Graph Transformers, sequence/temporal models, structured reasoning)
- ML infrastructure / data engineering for large-scale training + evaluation
Financial domain experience is a plus:
- Consumer banking, payments, investments, risk ops, or back office systems
- Or relevant coursework / education in financial systems / finops
(We do not expect candidates to have all of these. Deep strength in one area + strong ML fundamentals + strong engineering instincts is ideal.)
Hybrid (preferred): 1+ in-person days per week with teammates in SF Bay Area, Chicago, or Raleigh. Remote (OK): If remote, you’ll need working hours that comfortably support customer + internal meetings spanning US Pacific (PT), US East (ET), and European (CET) time zones
Onboarding: 1–2 weeks in person at our SF Bay Area HQ.
Travel: periodic travel for strategic customer workshops and deployments.
- Support and eventually lead 2–4 major customer engagements, delivering measurable business impact.
- Solve multiple challenging financial ML problems using rigorous evaluation and sound modeling choices.
- Ship at least one meaningful platform improvement driven by what you see on real customer datasets.
- Earn trust from customer technical teams and become their go-to person for ML strategy and execution.
- Partner with GTM to convert technical wins into production deployments and expansions.
This role offers a rare combination:
- Frontier ML on highly connected financial datasets where Graph Transformers can unlock step-change improvements.
- Field-to-core leverage: your learnings become shipped product capabilities, not one-off work.
- Ownership + speed: you’ll move fast, lead critical workstreams, and see direct impact.
- Career acceleration: deep technical work plus customer-facing technical leadership—the skill set that compounds quickly.
You’ll thrive in this role if you bring strong ML fundamentals, excellent instincts with people, and a drive to push yourself—and the technology—further than you ever have before.
The base pay range for this role is $125,000 – $185,000 per year.
Am I A Good Fit?
Get Personalized Job Insights.
Our AI-powered fit analysis compares your resume with a job listing so you know if your skills & experience align.
Success! Refresh the page to see how your skills align with this role.
The Company
What We Do
Democratizing AI on the Modern Data Stack! The team behind PyG (PyG.org) is working on a turn-key solution for AI over large scale data warehouses. We believe the future of ML is a seamless integration between modern cloud data warehouses and AI algorithms. Our ML infrastructure massively simplifies the training and deployment of ML models on complex data. With over 40,000 monthly downloads and nearly 13,000 Github stars, PyG is the ultimate platform for training and development of Graph Neural Network (GNN) architectures. GNNs -- one of the hottest areas of machine learning now -- are a class of deep learning models that generalize Transformer and CNN architectures and enable us to apply the power of deep learning to complex data. GNNs are unique in a sense that they can be applied to data of different shapes and modalities.
