• Define and drive the multi-year technical roadmap for the Core GenAI Platform, influencing adoption and alignment across several engineering teams • Lead architecture and design of large-scale, mission-critical systems for model inference, orchestration, and compliance • Identify and solve systemic technical risks that impact multiple teams and business-critical AI initiatives • Build high-leverage, generalizable platform components that meaningfully accelerate the productivity of engineers across Upstart • Mentor senior engineers (L5) and shape the growth of future technical leaders, fostering a culture of learning and excellence • Collaborate with VP-level and above leadership to influence business strategy with technical insight • Represent Upstart internally and externally as a thought leader in GenAI systems through executive reviews, cross-company forums, and industry venues • Establish operational standards for GenAI integrations, ensuring security, fairness, and reliability at scale

Junior Go/Typescript Engineer (AI-Assisted Development) About Lexmata Lexmata is a forward-thinking legal technology company building cutting-edge products that leverage AI/ML and modern cloud infrastructure to solve complex problems in the legal industry. We operate as a small, tight-knit remote team where every engineer has a direct impact on the product and the direction of the company. Position Overview We are looking for a Junior Engineer with real-world experience to join our small engineering team. This is not an entry-level position — we need someone who has already shipped code in a professional environment and is ready to hit the ground running with AI-assisted development workflows. You will work across our stack, contributing to Angular frontends, Go services, and cloud infrastructure on AWS, using tools like Claude Code and Cursor to accelerate your output. This role is ideal for an engineer who thrives in a startup environment: self-directed, resourceful, comfortable with ambiguity, and motivated by ownership rather than hand-holding. Key Responsibilities - Develop and maintain Angular frontend applications with a focus on performance and user experience - Build and extend Go microservices that power our document processing and AI pipelines - Write, test, and deploy containerized applications using Docker - Provision and manage cloud resources on AWS - Leverage AI-assisted coding tools (Claude Code, Cursor) daily to accelerate development and improve code quality - Participate in code reviews, architectural discussions, and sprint planning - Troubleshoot and resolve issues across the full stack independently - Collaborate asynchronously with a small remote team, communicating clearly and proactively with AI-assisted workflows - Adapt quickly to shifting priorities and new technical challenges 0

• Independently design and develop specialized LLM training protocols for AI Academy curriculum • Create detailed training datasets and fine-tuning strategies for AI tutors' instructional methods • Implement evaluation frameworks to assess AI tutor performance • Develop automated feedback systems to replace traditional assessment • Create knowledge bases for guiding hands-on exercises • Design problem-solving approaches that mimic expert guidance • Build systems for peer learning facilitation

• Define and evolve our infrastructure to allow for better ML and AI capabilities, with a focus on LLM-based and agentic systems. • Contribute to the development and expansion of our agentic AI framework powered by AWS Bedrock, enabling both internal tools and customer-facing features. • Identify, source, and refine datasets to allow tuning models, powering retrieval pipelines, or expanding agentic workflows. • Pre-process data by using techniques such as data cleaning, feature engineering, and transformation. • Train, evaluate, and deploy both LLM-based systems and traditional machine learning models into production. • Monitor, debug, and continuously improve deployed models and AI tools. • Support machine learning usage throughout the company, including selecting the right modeling approach for the use case (LLM vs. traditional ML). • Support the integration and use of LLMs, including approaches such as fine-tuning, prompt tuning, and retrieval-augmented generation (RAG), to improve accuracy.