About this role
Ancestry is seeking an exceptional and highly motivated Postdoctoral Research Fellow to join our AI Applied Science Content team. This fellowship is designed for a researcher at the intersection of industry-scale data and academic rigor. In this role, you will lead research at the forefront of Document Understanding, designing and implementing AI-Native agentic systems to unlock insights from billions of historical and genealogical records. Your work will focus on advancing the state of the art in autonomous multi-agent workflows and on transforming unstructured historical records into structured, searchable knowledge that connects customers to their family history.
What you will do:
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).
Architect Agentic Systems: Design and implement multi-agent workflows using frameworks like LangChain, LangGraph, CrewAI, AutoGen, AgentCore, Strands, Google ADK, A2A, etc. to automate complex multi-step reasoning tasks in historical document analysis and information extraction.
Analyze and Optimize Multi-Modal Models: Evaluate the performance of multi-modal AI & LLMs such as GPT, Gemini, Claude, Llama, and Qwen for zero-shot and few-shot scenarios in comprehensive document understanding.
Natural Language Processing (NLP): NER, Relation Extraction, Coreference Resolution, Entity Resolution, Knowledge Graphs (Neo4j), spaCy, NLTK, BERT.
Computer Vision (CV): Apply expertise using models like YOLO, Nougat, DONUT, OpenCV, etc. to perform layout analysis, identifying text blocks, headers, tables, and deeply nested lists.
Evaluation & Observability: Establish ensemble models and "LLM-as-a-Judge" frameworks, and use tools like Arize Phoenix, DeepEval, or RAGAS to monitor or hallucination, drift, and bias.
Development Productivity: Familiarity with "AI coding" workflows and usage of AI coding assistants such as Amazon Q, Cursor, Claude Code, and Kiro to accelerate development cycles.
Collaborate on Cloud Deployment: Partner with ML Ops to deploy datasets, models, and pipelines in cloud environments like AWS (S3, SageMaker, Bedrock, ECS, EKS) and GCP (Vertex AI, Gemini API).
Who You Are:
Ph.D. (recent graduate or near completion) in Computer Science, Data Science, Statistics, Linguistics, Engineering, or a related quantitative field with a strong research focus.
A strong record of academic publications in NLP, CV, or Agentic AI is preferred.
Specialization in AI & LLMs, including familiarity with foundational models such as GPT, Gemini, Qwen, Llama, Claude, etc.
Research in inference efficiency and optimization, potentially using vLLM, LoRA, QLoRA, and quantization approaches.
Familiar with embeddings, vector databases, and transformer models, with software development experience.
Strong proficiency in Python and relevant tools and libraries, including transformer models and multi-modal models.
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 are a plus.
Ability to clearly present complex technical solutions to both technical and non-technical stakeholders