Data Scientist, AI Data Foundations
Data Scientist, AI Data Foundations
About the Role
Reporting into the Data Engineering organization, the Data Scientist is responsible for designing and building the curated data structures that AI and ML applications consume across MeridianLink. You will own the vector stores behind our RAG systems, the feature store that powers model training and inference, and the graph databases that capture relationships across applicants, products, and decisions. You will also lead targeted data discovery work, surfacing hidden trends in our lending and account-opening data that inform both AI use cases and the broader business.
This is a hands-on, build-oriented role. You will not be primarily training large models — you will make sure the people training and serving models have high-quality, well-governed, well-engineered data to work with, and you will use your data science skills to validate that the data is fit for purpose.
What You Will Do
• Build and maintain vector stores for RAG: Design embedding pipelines, chunking strategies, indexing approaches, and refresh patterns for the vector stores powering retrieval-augmented generation across MeridianLink products.
• Own the feature store: Design, build, and operate feature store assets used for model training and online/offline inference, including feature definitions, freshness SLAs, lineage, point-in-time correctness, and reuse across teams.
• Design graph data structures: Build graph databases that model relationships between applicants, applications, products, lenders, decisions, and outcomes — and make them queryable for both AI use cases and analytical investigations.
• Lead data discovery: Profile our lending, deposit, and behavioral datasets to identify hidden trends, segments, anomalies, and potential model drivers; turn findings into actionable hypotheses for product, risk, and growth teams.
• Engineer for AI consumption: Build the curated, AI-ready datasets that downstream model builders, application engineers, and analysts rely on — with appropriate quality, documentation, and governance baked in.
• Evaluate retrieval and feature quality: Define and run evaluation frameworks for RAG retrieval quality, feature drift, embedding quality, and graph completeness; iterate based on what the metrics tell you.
• Partner with model builders: Work closely with ML engineers and applied scientists to make sure the data structures you build accelerate their work rather than slow it down.
• Champion responsible data use: Partner with governance, security, and compliance to ensure that AI-facing data assets respect data classification, customer consent, and regulatory boundaries from day one.
• Communicate findings: Translate discovery work into clear narratives — write-ups, notebooks, dashboards, and short presentations — that help non-technical stakeholders act on what the data is showing.
Required Qualifications
• 4–7 years of experience in a data science, ML engineering, or applied data role, with a meaningful portion of that time spent building data assets that other people's models or applications consumed.
• Hands-on experience designing and operating vector stores for RAG or semantic search, including embedding generation, chunking, indexing, and retrieval evaluation.
• Experience building or operating a feature store (e.g., Databricks Feature Store, Feast, or a custom internal platform), including offline training and online serving patterns and point-in-time correctness.
• Experience modeling and building graph data structures using Neo4j, TigerGraph, Azure Cosmos DB Gremlin, or similar graph databases — and writing graph queries to answer real questions.
• Strong proficiency in Python (pandas, NumPy, scikit-learn, PySpark) and SQL; comfortable working day-to-day in Databricks notebooks and jobs.
• Practical experience with embedding models and LLM tooling (e.g., Hugging Face transformers, OpenAI / Azure OpenAI APIs, LangChain or similar) in a production or near-production context.
• Demonstrated data discovery skills: profiling messy real-world datasets, surfacing non-obvious patterns, validating findings statistically, and explaining them clearly.
• Solid grounding in classical ML concepts — supervised vs. unsupervised learning, train/test discipline, leakage, evaluation metrics — even though you will not own model training day-to-day.
• Strong written and verbal communication skills; able to write up findings for both technical and business audiences.
Preferred Qualifications
• Experience working in a SaaS or FinTech environment, particularly with lending, deposit, credit, fraud, or KYC/AML data.
• Experience with Databricks-native AI/ML tooling: Databricks Vector Search, Databricks Feature Store, MLflow, and Unity Catalog.
• Familiarity with open-source vector databases such as pgvector, Pinecone, Weaviate, Chroma, or FAISS, and a clear point of view on when to use which.
• Experience with Microsoft Azure data and AI services (Azure OpenAI, Azure AI Search, ADLS Gen2).
• Experience evaluating RAG systems end-to-end (recall@k, faithfulness, answer quality, hallucination measurement).
• Exposure to graph algorithms (community detection, link prediction, centrality) applied to real business problems.
• Bachelor's or Master's degree in Computer Science, Statistics, Mathematics, Engineering, or a related quantitative field, or equivalent professional experience.
Our Data & AI Stack
• Lakehouse: Azure Databricks, Delta Lake, Unity Catalog, PySpark, SQL
• AI Data Foundations: Databricks Vector Search, Databricks Feature Store, MLflow
• Vector & Graph (current and exploratory): pgvector, Pinecone, Weaviate, FAISS; Neo4j, TigerGraph, Azure Cosmos DB (Gremlin)
• Cloud: Microsoft Azure (ADLS Gen2, Azure OpenAI, Azure AI Search, Event Hubs)
• AI Models and Agents: Databricks, AWS Bedrock, Azure ML
• Integration & Governance: Informatica Data Management Cloud (IDMC), Unity Catalog
As published by ashby
Name, Email, Resume
- Phone Number
- Are you eighteen years of age or older? yes / no
- May MeridianLink contact your CURRENT or MOST RECENT employer? yes / no
- May MeridianLink Contact your PAST employers? yes / no
- Have you ever been fired or asked to resign to avoid being fired from a job? yes / no
- Can you perform the essential functions of this job with or without reasonable accomodation? yes / no
- Will you now or in the future require sponsorship (i.e. H-1B visa ect.) to legally work in the United States? yes / no
- Location
- I understand that as permitted by law, MeridianLink will conduct an investigative consumer report. I understand the scope of this consumer report may include verification of my references, employment history, credit and indebtedness, criminal conviction history, educational and training background and other matters related to my suitability for employment. Upon timely written request to the personnel department of MeridianLink, the nature and scope of the report will be disclosed to me. If I am offered employment with MeridianLink, I understand that the offer will be conditioned on passing a pre-employment drug screen, timely submitting valid documentation that confirms my identity and authorization to work in the United States, and reading and agreeing to comply with the MeridianLink policies as stated in its Employee Handbook. I certify that the answers given by me in this employment application are true, correct and complete. I agree that the company shall not be liable, in any respect, if my employment is terminated because of misstatements or pertinent omissions made by me in this application. Moreover, I understand that all offers of employment are contingent upon passing the company's prescribed credit check and background screen. A copy of this form may be used as the original. The use of results from this form and/or tests will be used for prudent employment decisions. It is agreed and understood that completion of this application does not mean a job opening exists and in no way obligates the company to employ me. In the event of employment, I will comply with all company rules and regulations as established from time to time. I am willing to work all assigned overtime or other special work assignments as requested by the company. I also understand that MeridianLink, Inc. retains the right to amend, modify, add or delete any or all policies or procedures at its sole and absolute discretion. I understand that nothing contained herein is intended to create a contract between the company and me for either employment or the provision of any compensation or benefits. I hereby understand and acknowledge that any employment relationship with this Company is of an “At-Will” nature, which means that the Employee may resign at any time and the Employer may discharge Employee at any time, with or without notice, with or without cause. It is further understood that this ‘At-Will’ employment relationship may not be changed by any written document or by verbal agreement unless such change is specifically acknowledged in writing by an authorized Executive of this Company. During my employment with MeridianLink, Inc. and after my employment ends, I agree not to disclose any confidential or proprietary information regarding operating and trade secrets. CERTIFICATE OF APPLICANT: I certify that all statements made in this application and attachments are true, and I agree and understand that misstatements or omissions of any material fact may be cause for disqualification or dismissal from employment with MeridianLink choose one
- Please let us know where you heard about this job opening? choose one
- If Referral please add name below. optional
- If above question is "Other" please add the source below. written answer · optional