ML Engineer Intern

About Kita

In emerging markets, credit infrastructure is broken. Credit bureaus are unreliable, open finance is limited, and most financial data lives in messy documents. As a result, millions of individuals and businesses are locked out of credit because lenders can’t properly assess them.

Kita is building the infrastructure to fix this. We turn messy financial documents into structured, fraud-checked signals that lenders use to make underwriting decisions. We work with customers ranging from fintechs to enterprise banks, and build closely with them on the ground — during the YC batch, we spent time in Manila, Singapore, Jakarta, and Mexico City.

We’re a Stanford AI team backed by Y Combinator, top funds, and leading angels across Silicon Valley and Southeast Asia. During the YC batch, we grew ~40% week-over-week with customers across three continents. Our CTO was ranked #1 in Stanford CS in 2025.

The opportunity

Kita is looking for an exceptional Applied ML Intern to work directly with the founding team on some of the hardest problems in lending, fraud detection, and document intelligence.

We want someone highly technical, extremely fast, and excited to take on complex financial, credit, ML problems — from model prototyping and evaluation to data pipelines, backtesting, vision systems, and production-facing experiments.

You’ll work across machine learning, data science, computer vision, and product engineering to help build the intelligence layer behind Kita’s products. That includes figuring out which signals in messy financial documents actually predict repayment and fraud, designing evaluation systems for high-stakes underwriting workflows, and helping turn raw models into systems that customers can trust.

What you’ll work on

• Build and test ML systems across document intelligence, fraud detection, credit risk, and underwriting
• Run backtests on historical financial lending and document data to identify predictive signals
• Prototype models in computer vision, multimodal, and LLM/VLM workflows
• Design evaluation frameworks, error analysis tools, and benchmarking systems for real customer use cases
• Collaborate directly with founders, engineers, and customers to turn messy real-world problems into working systems

Strong candidates will likely have:

• Strong technical background: foundations in ML, data science, or applied math, with the ability to move from first-principles modeling to high-quality implementation quickly
• Extremely strong builder who can ship end-to-end systems fast
• Experience benchmarking, evaluating, and debugging ML systems on messy, noisy, real-world data
• Comfortable operating across a broad range of technical problems, rather than within a narrowly scoped research or internship lane

Especially exciting backgrounds include:

• Prior work in fintech, credit, fraud, risk, underwriting, or financial data
• Experience with computer vision, multimodal models, voice models
• Experience with backtesting, feature selection, model calibration, or decision systems
• Research experience in applied ML, RL, vision, or LLM/VLM systems