Agentic Dream is a company focused on transforming Standard Operating Procedures into executable agent logic. They are seeking a Technical Business Analyst who will bridge the gap between business intent and technical specifications, ensuring that complex business processes are accurately translated into actionable requirements for data and AI projects.

Responsibilities:

Deconstruct complex business processes and user stories into EARS-formatted requirements (Easy Approach to Requirements Syntax) that are unambiguous, testable, and traceable
Author Gherkin scenarios (Given/When/Then) that define the "Definition of Done" — including for non-deterministic agent behavior, where you specify the logic path, not just the output
Maintain the Master Contextual Package (MCP) traceability: every requirement maps to a business need and to the agent/pipeline logic that implements it
Build and maintain data dictionaries, schema specifications, and source-to-target mappings for the Bronze → Silver → Gold pipeline and the PostgreSQL schema registry
Define field-level validation rules, tolerances, and alias dictionaries (e.g., "Qty Shipped" → quantity) that drive the Schema Matching and validation MCP servers
Profile source data, quantify data-quality issues, and write the acceptance criteria for data-quality gates and cross-field validation (e.g., line items sum to total)
Translate business policy into structured, machine-consumable business-rule definitions (discount limits, tax rules, workflow steps) and the guardrail specs that enforce them ("Block discount > 20%")
Maintain the knowledge-pack catalogs (business rules, guardrails, knowledge base) as living, versioned specifications
Map current-state and future-state quote-to-install and document-processing workflows, identifying automation opportunities, HITL (Human-in-the-Loop) decision points, and exception paths
Define confidence-based routing rules with stakeholders (≥90% auto-approve, 60–89% review, <60% Expert-in-the-Loop) and the SLAs around them
Run requirements workshops and translate technical complexity (why an automation is risky, costly, or non-deterministic) into language non-technical stakeholders can act on
Own UAT planning and acceptance: build test cases from the Gherkin specs, coordinate sign-off, and verify delivered work matches the spec
Produce operational reporting specs — defining the metrics (throughput, accuracy, cost-per-document, net savings) and dashboards the business needs
Run discovery sessions with business stakeholders and engineers; produce EARS requirements, Gherkin acceptance criteria, and data mappings
Write SQL queries to profile and validate data, quantify edge cases, and back requirements with evidence (not assumptions)
Maintain the requirements backlog and traceability matrix in Jira (project SDB) and Notion
Partner with Data Engineers on schema/registry specs and source-to-target mappings
Partner with AI Engineers to define agent behavior, tool contracts, and the "Definition of Done" for probabilistic outputs
Work with the PM team to scope sprints, estimate, and sequence delivery
Build and execute UAT test cases from the specs; document defects with clear reproduction steps
Validate that delivered pipelines and agents meet acceptance criteria before release
Maintain decision logs and ADR-style records of requirement changes

Requirements:

5+ years as a Business Analyst / Technical BA, with 3+ years on data-intensive projects (data platforms, integrations, analytics, or document/IDP processing)
Strong SQL — able to independently profile data, validate hypotheses, and write moderately complex queries against PostgreSQL or similar
Demonstrated requirements engineering skill — user-story decomposition, acceptance criteria, and at least one formal technique (EARS, BDD/Gherkin, use cases, or equivalent)
Data modeling literacy — able to read and contribute to data dictionaries, schema specs, ER diagrams, and source-to-target mappings. Understands normalized vs. denormalized models conceptually
Process modeling — BPMN or equivalent; mapping current/future-state workflows and identifying automation and exception paths
Familiarity with the modern data & AI stack — understands what RAG, vector databases, APIs/microservices, and LLM-based automation are and where they fit (you don't have to build them, but you must spec for them)
Tooling — Jira, Confluence/Notion, and diagramming tools; comfortable maintaining traceability matrices and living documentation
English proficiency: B2+ required (C1 preferred). You'll facilitate workshops, write specs, and translate between business and technical audiences daily
Experience specifying agentic / LLM-driven systems — defining 'Definition of Done' for non-deterministic outputs and writing guardrail/policy requirements
Hands-on with document/IDP projects — POs, ACKs, Invoices, schema/alias matching, confidence scoring, and HITL review workflows
Familiarity with Spec-Driven Development (SDD) and the Model Context Protocol (MCP) concept
Experience writing YAML/JSON business-rule definitions or configuration-as-spec
Light scripting (Python) for data profiling and analysis
BI/reporting spec experience (QuickSight, Power BI, Metabase)
Background in commercial furniture, logistics, distribution, or manufacturing operations
Domain exposure to quote-to-install / order management processes

Business Analyst- W2

About this role