NLP / LLM Data Scientist

Dandelion Health AI; New York (NY); Remote; USD 120,000 - 180,000; Full time

Benefits offered by this job

Remote work

Wellness benefits

Professional development

Collegial work environment

Academic problem solving culture

Startup speed and flexibility

Job summary

Dandelion Health AI is seeking a healthcare data scientist to build NLP/LLM pipelines and curate datasets from multimodal clinical data. You will work with structured and unstructured data, including EMRs, notes, and images, to support AI model development and real-world evidence analyses.

You will collaborate with the Data Science Manager to deliver high-quality, compliant data products and insights for healthcare clients in a fast-paced startup environment.

Qualifications

• Advanced degree in a quantitative field or a B.S. with at least 5 years of experience.
• 2+ years of data science and ML experience with unstructured/semi-structured data.
• Fluency in Python and SQL and ML/NLP libraries (PyTorch, TensorFlow, HuggingFace).
• Experience with applying LLM techniques to real-world health data encouraged.
• Strong writing, communication, and project management skills.

Responsibilities

• Develop NLP/LLM pipelines to abstract labels from text-based healthcare data.
• Query EMRs, semi-structured reports, and notes to build high-quality datasets.
• Own data extraction, wrangling, labeling and QC for analytical datasets.
• Stay current on NLP and generative AI methods and apply them as appropriate.
• Support design, testing, validation, and merging of multimodal data structures.
• Develop code and documentation for HIPAA-compliant data products on time.
• Troubleshoot problems using subject matter knowledge and critical thinking.
• Ensure data accuracy, integrity and validity across analyses.
• Provide support to the technical product team for data offerings.
• Summarize methods and findings into clear explanations for diverse audiences.

Skills

Python

SQL

NLP

Data pipelines

HIPAA knowledge

Education

Advanced degree in quantitative field

BS with 5 years experience

Tools

PyTorch

TensorFlow

HuggingFace

AWS

Job description

Our Team

Dandelion Health was founded in 2020 by experts in health tech, hospital systems, academia, and clinical AI. We are building the world’s largest AI training and clinical development platform. Today, we pride ourselves on our ability to make data access as easy as possible for AI developers, pharma, and medical devices, while raising the bar for patient safety and data quality. Tomorrow, we will be the place where any healthcare organization can go to build a responsible clinical AI product. Our culture is all about learning from data and improving, so we can help our clients improve health through AI. Meet the rest of our team here.

Our Data

We partner with health systems to safely and ethically make their de-identified patient data available to AI developers. Currently, the data is acquired from Sharp HealthCare, Sanford Health, and Texas Health Resources – with two additional U.S. health systems joining soon.

We have clinical data dating back to July 1, 2016. This data represents over 10 million patients and includes but is not limited to:

• Structured data (e.g., 100% of the EMR, including some claims)

• Unstructured text (e.g., clinical notes, radiology reports)

• Images (e.g., DICOM, pathology)

• Video

• Waveforms

• Continuous streaming monitoring data

Your Role

You are a healthcare data scientist who knows your way around clinical and text-based data. Your primary responsibility will be to leverage and build upon large language models (LLM) and other ML-based approaches for meaningful data abstraction from both unstructured and structured healthcare data. You will join a team of data scientists who own the creation and maintenance of AI-ready datasets for clients in our environment. You will work across the organization to curate datasets by identifying patient subpopulations or disease cohorts, developing methodology to abstract information from multimodal healthcare data sources to support patient phenotyping, and pooling data to enable rapid exploratory AI/ML analyses, model experimentation and/or model validation. You will use your data expertise, programming abilities, and critical thinking skills to support our technical product team, and develop your own analyses to derive insights and enhance our datasets based on the use case. Your team’s ultimate goal is to deliver the highest-quality data possible to our clients, who are building products that improve patient health. You will report to the Data Science Manager, under the Head of Data.

Responsibilities

Your day-to-day responsibilities will include the following:

• Develop Natural Language Processing (NLP), Large Language Model (LLM) and other ML-based pipelines to abstract relevant labels from text-based healthcare data and store them in scalable data models;

• Query complex source systems in a range of health data sources (e.g., EMRs, semi-structured reports, free-text clinical provider notes) to identify key data elements and create and enrich high-quality datasets for real-world evidence analyses and training AI algorithms;

• Own data extraction, wrangling, labeling and QC tasks to create analytical datasets that include abstracted clinical concepts and provide a range of solutions to support customers’ AI activities;

• Stay current on the latest in applied NLP and generative AI methods and proactively leverage these technologies where applicable;

• Support the design, testing, validation, analysis, and merging of multimodal data structures from a wide variety of source systems;

• Develop code and documentation to deliver high-quality and HIPAA-compliant data products on time to customers;

• Identify and resolve problems using your knowledge, background, and troubleshooting skills;

• Ensure accuracy, data integrity, and validity of data and analysis in all work;

• Provide support for technical product team to advance development of the suite of data-related product offerings;

• Summarize the complexity of abstraction methods, findings and recommendations into clear explanations and presentations for internal and external audiences that have a varying range of technical and clinical experience;

You are not afraid to dig into massive, confusing, disorganized new datasets and get them under control. You are excited to learn new environments, languages, and skills. This is a small, early stage company with enormous ambitions and everyone pitches in across the team.

Qualifications

• Advanced degree in a quantitative field (ex. Data Science, Biomedical Informatics, Computer Science, Biostatistics), or B.S. with at least 5 years of professional experience

• At least 2 years of data science and machine learning experience, including building pipelines to extract and curate unstructured and semi-structured data by applying advanced machine learning and AI techniques. Prior experience with clinical and healthcare data is a strong bonus.

• Fluency in Python and SQL, including fluency with ML/NLP libraries (PyTorch, Tensorflow, HuggingFace, etc.)

• Familiarity with using modern applied LLM techniques on real-world data

• Strong technical writing, editing, and communication skills, along with a collaborative mindset

• Excellent organizational skills with an ability to embrace change and effectively manage multiple projects and consistently plan work to meet deadlines

• Experience working in or with startups is a plus

Technical Experiences and Skills

We don’t expect anyone to have all of the following skills or experiences, but we do seek candidates who are interested in growing their skill sets and working with healthcare data in all its glorious complexity. The Data Team works closely with our Engineering Team to put our work into production and meet client needs.

• Git and version control

• Familiarity with encryption methods

• Prior experience querying EDWs or databases and creating reports or analytics for healthcare data

• Familiarity with the data aspects of electronic medical records, ex. Epic, Cerner, Allscripts

• Any medical ontology experience

• Any experience working with DICOM or other imaging modalities

• Experience with AWS

• Experience with publishing work in peer-reviewed journals

Nature of our work

Our work is fast paced and iterative. We are growing, and we want to support our team members to grow in their skills as well. We are building a team that approaches problems with a diversity of perspectives, values experimentation, and refining our approach based on that experimentation. We work with the full spectrum of healthcare data from tabular data, videos, images, waveforms, etc. If a health system collects it, we might work with it!

If this looks like a partial fit, please reach out, we would love to share more about the work we do for you to understand if it would be a good fit for you.

There is occasional travel for in-person company working days on roughly a quarterly basis.

Team Benefits

• Remote work and flexible hours. Availability needed for meetings, which we try to keep to a healthy minimum

• Complete wellness benefits including healthcare, dental, vision, PTO, sick days and more. Ask for details

• Professional development days to build your skills

• Collegial work environment

• Academic bent towards inquiry and problem solving but start-up speed and flexibility

• Great balance of focus time to work on projects but easy to access team members to discuss issues and work collaboratively

• Dandelion is a mission-driven company that is focused on improving patient care