Summer Internship - AI Researcher
Aledade
Primary Duties
- Schema & Protocol Architecture (25%): Design a unified request/response schema that abstracts variations in proprietary EHR APIs, enabling downstream AI applications to request patient context agnostically.
- Agentic Fallback Routing (25%): Develop the logic to detect incomplete or failed API requests and deploy a browser-use agent to locate and extract the missing context via the EHR's web interface.
- LLM Data Normalization (25%): Build a reasoning layer utilizing LLMs/VLMs to process unstructured documents retrieved by the agent, extract required clinical elements, and map them to the UECP schema.
- Performance & Reliability Evaluation (25%): Establish an evaluation framework to measure the operational tradeoffs between API retrieval and agentic fallback. Design caching strategies to mitigate latency, and implement automated LLM evaluation pipelines (e.g., LLM-as-a-judge) to assess extraction accuracy and clinical safety.
Minimum Qualifications
- Education: Currently pursuing a Master’s or PhD in Computer Science, Applied AI, Software Engineering, Health Systems Engineering, or a closely related discipline.
- Programming: Strong backend software engineering skills, primarily in Python, with a solid foundation in data structures, system architecture, and JSON schema design.
- Web Automation: Experience with web scraping, DOM manipulation, and browser automation frameworks (e.g., Playwright, Puppeteer, Selenium).
- AI/Machine Learning: Practical experience integrating LLMs and Vision-Language Models (VLMs) for unstructured data extraction and reasoning.
Preferred Knowledge, Skills, or Abilities
Agentic Frameworks: Proven experience or deep academic interest in building autonomous, browser-use agents, semantic routing, and fallback logic (e.g., LangChain, AutoGPT, or custom reasoning loops).
Healthcare Interoperability: Understanding of standard healthcare data exchange protocols (like HL7 FHIR, SMART on FHIR), EHR API ecosystems, and clinical coding models like Hierarchical Condition Categories (HCC).
System Optimization: Ability to evaluate and optimize the operational tradeoffs of AI systems, specifically balancing latency, caching strategies, and extraction accuracy in real-time environments.
AI-Assisted Engineering: Proficiency in using AI coding tools (e.g., Claude Code, Cursor) to quickly prototype and bypass boilerplate engineering tasks, keeping the focus on core routing architecture.
Research & Autonomy: High tolerance for ambiguity and the ability to independently research, test, and architect fault-tolerant systems in highly fragmented and unpredictable software ecosystems. Strong technical writing skills for potential academic publication.