Machine Learning Engineer Intern

Machine Learning Engineering Intern

Branch's ML based underwriting models are core to our business and directly drive business value. Every credit decision, fraud signal, and credit scoring model we deploy has a direct impact on real customers' access to capital. We are looking for engineers who want to grow with us, take deep ownership, and have a genuine impact on how ML is built and deployed in production. We are hiring ML engineering interns to join our ML team in India. This team owns our core credit models, infrastructure, and the tooling that powers them. They also actively research and develop what comes next for ML at Branch for real credit and lending problems.

As a company, we are passionate about our customers, fearless in the face of barriers, and driven by data. As an engineering team, we value bottom-up innovation and decentralized decision-making: We believe the best ideas can come from anyone in the company, and we are working hard to create an environment where everyone feels empowered to propose solutions to the challenges we face. We are looking for individuals who thrive in a fast-moving, innovative, and customer-focused setting.

Here are some things you'll do

• Build, maintain, and scale ML infrastructure end-to-end: feature computation pipelines serving real-time predictions at millions of requests per day, and training and deployment pipelines spanning classical and cutting-edge models.
• Build observability and drift monitoring to detect and respond to model degradation in production.
• Evaluate and address selection bias in credit models using techniques from reject inference to deep generative approaches.
• Explore using LLMs/agentic AI to generate features from raw data, automate segment analysis, inform credit and policy exploration, and more.
• Collaborate with credit underwriting, product, and backend teams to build signals from structured and unstructured data that identify creditworthy borrowers and power our underwriting models.

Minimum Requirements

• Have some experience building ML applications that are more than just toy projects. Applications that have real users or an open source project would be a great plus.
• Strong skills in building machine learning models using both structured and unstructured data.
• Strong Python proficiency, including ML libraries (XGBoost, scikit-learn, pandas, numpy) and software engineering fundamentals.
• Experience with SQL and structured data, able to write non-trivial queries for feature extraction and analysis.

Nice to have

• GenAI/LLM experience
  1. Prompt engineering, fine-tuning, or building LLM-powered workflows
  2. Familiarity with agentic system design and using structured LLM outputs.
• Credit risk or lending domain experience
  1. Credit risk modeling or loan underwriting in any market.
  2. Leveraging alternative data sources as predictive signals in emerging markets.
• Experience with Rust or compiled languages
• Curious about new technologies and interested in learning different parts of the technology stack beyond ML when needed.
• Demonstrated fluency with modern AI-assisted development workflows and a point of view on their effectiveness for various use cases.
• Clear, structured written and verbal communication, comfortable operating async effectively.

Core Engineering Expectations

The following describes the traits we expect in our engineers, and reflects instincts developed over years of shipping real software. If you are early in your career, we do not expect you to demonstrate all of it yet — but we do expect a genuine appreciation for why these qualities matter, an honest sense of where you stand against them, and the drive to close the gap.

Engineering craft and judgment

• You balance technical excellence with business and consumer needs. You research and seek opinions to identify the best solution for a problem, not just the first one that comes to mind. You intentionally weigh technical debt, ensuring it is paid down before it compounds.
• You have a deep understanding of web fundamentals, data modeling, and distributed systems. You apply proven design patterns to build modular, scalable solutions and ensure that security, performance, and observability are baked into your work from the outset.
• You use automated testing as both a blueprint for clean architecture and a safety net for refactoring. You believe that well-designed code should be inherently testable, and you hold all contributions, whether human or AI generated, to the same standard of quality and maintainability.
• You are comfortable navigating and evaluating unfamiliar codebases. You can integrate into and evolve existing architectures with care, researching thoroughly and seeking peer feedback to ensure your changes improve the system without breaking it.
• You understand the full software development lifecycle. You can break down ambiguous projects into granular, estimable tasks that can be delegated effectively, ensuring a predictable and transparent path to delivery. You treat your delivery commitments as something the rest of the team plans against.
• You consider the ripple effects of your work. You think critically about how your choices impact the end user, the business, and the team's future velocity, and you have the drive to master the domain knowledge required to make those calls.
• You partner with product, design, and other functions to shape the work, not just receive it. You push back when the spec is wrong, ask the question that reframes the problem, and share ownership of whether the thing was worth building in the first place.

AI proficiency

• AI fluency is table stakes. You use AI to amplify your logic, not to replace it. You decompose ambiguous problems into clear steps and dependencies before prompting, ensuring you drive the architecture rather than allowing the AI to improvise a structure that compounds errors at high speed.
• You hold AI-generated output to the same rigorous standards as human written code. You have the intuition to catch hallucinations of correctness, e.g. elegant looking code that fails subtly or tests that pass without actually asserting logic, and you take total accountability for the final implementation. Your trust is calibrated through experience; you can clearly articulate your technical reasoning whenever you choose to override an AI's suggestion.
• You manage tool context deliberately to avoid "model noise" and recognize the point of diminishing returns. You know exactly when to leverage a prompt to accelerate your workflow and when to step back to solve a high-complexity problem through deep, independent thought.
• You proactively mitigate risks inherent to agentic workflows, including prompt injection, data leakage, and secret management. You treat AI-era security as a fundamental part of the engineering process, ensuring that "agentic" doesn't mean "unsecured."
• You aggressively automate the mundane so your attention is reserved for the complex logic and design decisions that actually need a human.
• You stay current with the evolving AI landscape and understand the underlying mechanics of the models and tools you use. You continuously raise your own productivity ceiling as the tools improve.

Operating principles

• You make high-velocity, high-quality decisions. You move fast where the cost of being wrong is low and reversible, slow down where it isn't, and can articulate which mode you're in and why.
• You are highly entrepreneurial and manage your own time, energy, and quality without oversight. You take the initiative to solve problems before they arise; nobody has to ask you twice, and often, nobody has to ask at all.
• You are inherently curious and a high-velocity learner who handles setbacks without losing momentum. You possess the grit to keep moving when things go sideways and the judgment to know exactly when to pull in help to stay on track.
• You have an uncompromising attention to detail. You catch the inconsistent behavior, the edge case bug, and the green test suite that doesn't actually verify anything, maintaining a high bar for the small things that define quality.
• You simplify the complex. You write with the clarity of someone who expects to be read and ask the sharp, uncomfortable questions that often reveal hidden project risks.
• You treat engineering as a team sport. You make those around you better through candid, helpful feedback, active listening, and a genuine commitment to helping your peers grow.
• You're intellectually honest. You change your mind when you're wrong, credit others when they're right, and never defend a position or a design just because it's yours.
• You operate with integrity. You make the right call when no one is watching, raise concerns early when something feels off, and treat the trust placed in you — by users, teammates, and the business — as