AI Architect (Generative AI & Databricks)
About this role
About Entrada AI
Entrada AI is a specialized consulting partner and a strategic portfolio company of Databricks Ventures. We were recently named the Genie Partner of the Year (2026) for our work deploying Databricks Genie at enterprise scale—bridging the gap between AI ambition and the trusted, governed data required for accurate responses. With over 150 Databricks projects delivered, we unlock self-service analytics for Fortune 500 leaders.
Databricks invested in us because of our technical excellence, placing us in the "inner circle" of the ecosystem. For our engineers, this means direct access to product roadmaps, private previews, and the teams building the platform. You will join a team of industry veterans who value clean architecture over quick fixes. We don't just maintain pipelines; we solve complex architectural challenges.
About the Role
Entrada AI, Inc. is seeking an experienced AI Architect (Generative AI & Databricks) to join our growing team of consultants to lead the design and implementation of an enterprise GenAI platform for a marquee client.
In this role, you will architect an intelligent ingestion pipeline and an Anti-Hallucination & Zero-Inference Architecture designed to automate the triage, parameter extraction, and staging of incoming customer commercial pricing requests. You will bridge advanced LLM orchestration with production Lakehouse patterns, replacing manual data re-keying with high-throughput unstructured parsing, automated guardrail verification, an interactive Human-in-the-Loop (HITL) review cockpit built on Databricks Apps.
Key Responsibilities
Design and oversee the end-to-end GenAI Lakehouse architecture on Databricks, transforming unstructured email threads and complex attachments into fully verified, ERP-ready tabular data.
Architect secure, incremental ingestion pipelines via Microsoft Graph API and PySpark parsing engines to process multi-page PDFs, Excel workbooks, and HTML bodies directly into Unity Catalog Volumes and Bronze Delta tables.
Lead the development of deterministic extraction workflows on Databricks Model Serving (Claude/GPT), enforcing strict Pydantic schemas to accurately extract commercial attributes without arbitrary prose.
Implement a 3-layer anti-hallucination verification engine combining strict schema constraints, mandatory source citations, and PySpark string-assertion to ensure zero hallucinated data reaches production.
Architect and deploy an interactive Human-in-the-Loop (HITL) review application natively on Databricks Apps (Streamlit/Dash) featuring side-by-side reconciliation, manual override capabilities, and secure SSO/OBO authentication.
Model Gold-layer pricing staging tables and design automated batch feeds for downstream ERP consumption via Databricks SQL Warehouses and REST APIs.
Establish best practices for Unity Catalog governance, data lineage, error-quarantine routing, and system security across client engineering teams.
Create and maintain comprehensive architectural documentation, data models, and interface specifications while mentoring client teams.
Requirements
Bachelor’s or Master’s degree in Computer Science, Artificial Intelligence, Data Engineering, or a related field.
7+ years of experience in data engineering, data architecture, or applied machine learning systems.
4+ years of hands-on experience designing and deploying solutions on the Databricks Lakehouse Platform (Unity Catalog, Delta Lake, Databricks Model Serving, and Databricks SQL).
Demonstrated experience as an Applied GenAI / LLMOps Engineer/Architect, specifically building structured extraction pipelines using function calling, Pydantic validation, and custom guardrail/citation-verification systems.
Proven track record parsing dirty, semi-structured, and unstructured business documents using Python and PySpark libraries (PyMuPDF, pdfplumber, OpenPyXL, BeautifulSoup4).
Hands-on experience developing and deploying interactive data applications using Streamlit or Dash, with specific experience deploying containerized apps natively inside Databricks Apps.
Strong understanding of enterprise identity and API integrations, including Microsoft Graph API, OAuth 2.0 client credential flows, Microsoft Entra ID (Azure AD), and Databricks Secret Scopes.
Solid background in data modeling, governance, and auditability inside Unity Catalog (handling Volumes, Delta Tables, ACLs, and row/column-level security).
Experience or working familiarity with downstream enterprise ERP integration patterns.
Strong problem-solving and analytical skills, with the ability to design resilient, production-ready systems that prevent hallucinated data from touching financial systems of record.
Excellent communication, consulting, and client-facing skills, with the ability to articulate complex GenAI concepts clearly to both technical teams and executive business stakeholders.
The Offer
Cooperation Models:
Poland: Available via Employment Contract (UoP) or B2B.
Romania/ Greece: Available via B2B only.
100% Remote: Full flexibility to work from anywhere in Poland/ Romania / Greece
High-End Tech: Apple MacBook Air M4 15" provided to all engineers.
Referral Bonus: Bonus for bringing other top-tier engineers to the team.
-
Professional Growth:
Certification Support: Coverage for all Databricks technical certifications.
Industry Leadership: Support in reaching the highest tiers of Databricks expertise (such as the Champion program) tailored to your specific career track.
Personal Branding: Opportunities to present at global industry conferences and contribute to technical thought leadership.
Expert Mentorship: Direct collaboration with Databricks MVPs and core product teams, giving you a front-row seat to the platform's evolution.
Recruitment Process
Introductory Call (20 min): Short conversation with our Recruiter to discuss your background and expectations.
Technical Interview (60 min): Deep dive into your technical skills with our engineering team.
Optional Client Interview: Required only in specific cases.
Decision & Offer: We aim to close the process and provide feedback efficiently.
Originally posted on Himalayas
