Data Science - Agentic AI, Document Understanding Co-op
Ancestry is a human-centered company that connects people with their family history. They are seeking a highly motivated Agentic AI, Document Understanding Co-op to design and implement AI systems that extract and organize information from historical records, working closely with engineering teams to optimize and deploy solutions. Responsibilities Innovate with State-of-the-Art AI: Implement cutting-edge AI solutions for key Document Understanding tasks such as OCR/HTR, transcription, Named Entity Recognition (NER), Relation Extraction (RE), Coreference Resolution, Summarization, and Knowledge Graphs working with diverse genealogical and historical collections spanning newspapers, city directories, family history books, and vital records (i.e., birth, marriage, & death records) Analyze and Optimize Multi-Modal Models: Evaluate the performance of multi-modal models in zero-shot and few-shot learning scenarios for comprehensive document understanding Architect Agentic Systems: Design and implement multi-agent workflows using frameworks like LangChain, LangGraph, CrewAI, or AutoGen to automate complex multi-step reasoning tasks in historical document analysis Evaluation & Observability: Establish 'LLM-as-a-Judge' frameworks and use tools like Arize Phoenix, DeepEval, or RAGAS to monitor for hallucination, drift, and bias Collaborate on Cloud Deployment: Partner closely with ML Ops and Data Science Engineers to seamlessly deploy datasets, models, and pipelines in cloud environments Communicate Insights Effectively: Clearly and confidently present your findings, deliverables, and proposed solutions to technical and non-technical audiences, including teams, stakeholders, and executives Skills Currently pursuing an advanced degree (Master's or PhD) in Computer Science, Data Science, Statistics, Mathematics, Linguistics, Engineering or related quantitative field with a strong data focus Specialization in AI & LLMs including familiarity with foundational models such as GPT, Gemini, Qwen, Llama, Claude, etc Experience with inference optimization, vLLM, LoRA, QLoRA, quantization, etc Familiar with embeddings, vector databases, transformer models, with software development experience Strong proficiency in Python and relevant tools and libraries, including transformer models, multi-modal models, and general NLP (e.g., Hugging Face Transformers, agentic frameworks and workflows, LangChain, LangGraph, CrewAI, AgentCore) Master's or PhD preferred in Computer Science, Data Science, Statistics, Mathematics, Linguistics, Engineering or related quantitative field with a strong data focus Familiarity with cloud platforms and related AI/ML services such as Google Cloud Platform, GCP, Gemini API, Vertex AI, AWS EC2, S3, SageMaker, Model Registry, and Bedrock Company Overview Ancestry is a web-based platform that helps its users to create their own family tree and help them preserve and share their family history. It was founded in 1983, and is headquartered in Lehi, Utah, USA, with a workforce of 1001-5000 employees. Its website is Company H1B Sponsorship Ancestry has a track record of offering H1B sponsorships, with 61 in 2025, 60 in 2024, 65 in 2023, 99 in 2022, 60 in 2021, 47 in 2020. Please note that this does not guarantee sponsorship for this specific role.