AI/ML Engineer

ABOUT KEEPERWe're a YC startup that combines great UX and heavy-lifting machine learning to deliver a whole new kind of tax filing experience.We have:40K paying users ($199 to $399 per year)20K reviews averaging 4.8 stars on the App Store, Play Store, and TrustPilot20 full-time employees, $15M raised (Series A)ABOUT THE ROLEWe're looking for an AI/ML engineer to build user-friendly features powered by cutting-edge tools. Here are some examples of projects we're currently working on:Improving our state-of-the-art transformer-powered tax form parserLeveling up our industry-leading AI tax assistantIterating on our core ML models: an NLP model to generate human-readable transaction descriptions, and a classification model to predict which transactions are tax-deductibleWHAT WE'RE LOOKING FORA successful applicant for this position will be:Deeply experienced in modern ML and AI systems, including OCR, RAG, LLM fine-tuning, prompt engineering, and AI agentsHighly proficient in Python, with working knowledge of GCP infrastructure and modern ML tooling (e.g. Pandas, HuggingFace, Weights & Biases)Intensely product-minded, with a focus on translating models into thoughtful, borderline magical featuresRapidly iterative, able to design smart tests, analyze outcomes, and refine features quicklyOwnership-oriented, taking end-to-end responsibility for code quality, feature reliability, and production issues when they ariseCollaborative and generous, willing and able to partner closely with cross‐functional teammatesPassionate about the potential of AI to build great user experiences; experience in the tax domain is a plus but not requiredLOCATIONWe have a mix of remote and in-office employees. For this role, we are looking for someone who wants to work one or more days per week in our SF office (Financial District).VISA REQUIREMENTSApplicants must be a U.S. citizen, green card holder, or H‐1B visa holder (transfer). We do not support F-1 OPT or STEM-OPT.Other visa classifications (e.g., TN, E‐3, O‐1) may be considered on a case‐by‐case basis.#J-18808-Ljbffr