--- title: "An Introduction to Data Roles - Snowpack Data" description: "A practical guide to which data roles are fungible, and where the new AI-era titles actually fold in" url: https://snowpack-data.com/blog/data-role-fungibility --- [Blog](https://snowpack-data.com/blog) # An Introduction to Data Roles A practical guide to which data roles are fungible, and where the new AI-era titles actually fold in [Kevin Koenitzer](https://www.linkedin.com/in/kevin-koenitzer-b00013a8/) · November 10, 2026 · 20 min read With the number of different data roles open on the market today, it’s difficult for an outside individual to figure out which roles are fungible across one-another and which are not. We wrote this article for non-data folks to resolve confusion around the roles in data, where they overlap, and provide a starting point for understanding how to build and scale a data organization from a staffing perspective. ## **The 4 core data roles** Though there are at least 10 different roles in a data practice, all of those roles are effectively related-to, or a derivative of one of the four core roles in data: Data Analyst, Data Engineer, Data Scientist and ML Engineer. Data Engineer Data Analyst Data Scientist ML Engineer 1 **Understands**Knows what it is and when it's needed. Wouldn't be handed the work. 2 **Can contribute**Does pieces of it, with review from someone stronger. 3 **Can deliver**Ships sound work alone. Not the person who sets the standard. 4 **Expert**The team's go-to. Handles the hard cases. 5 **Master of craft**Sets the standard and teaches it. Across these four core roles and their derivative counterparts there is a certain degree of fungibility, but there is little overlap between the roles themselves. As organizations expand in size and complexity, new roles adjacent to the core roles above develop. These are roles that at large scale are assigned to an individual or group, but at a small scale may be folded into the original four. A common pitfall in hiring and staffing data teams starts with hiring a core role that is not aligned with the actual needs of the business. Most often this trap manifests as a company hiring a data scientist or data analyst without any existing infrastructure for them to work from, and then expecting them to either build the infrastructure themselves, or produce production-scale outputs on top of messy data. The chart below shows how these roles relate to one another, and which ones fold into the core role at smaller scales. The X axis describes where each role sits in terms of relative business/domain expertise, and the Y axis describes where each role sits in terms of relative technical competency. Everyone wants to hire the unicorn person who can build everything from scratch across every discipline, and translate findings into actionable tools and or decisions for the business to take. In general though, there is an inverse relationship between technical expertise and business knowledge because data professionals that spend time building back-end systems rarely get the chance to develop their interpersonal communication skills, or learn about the business, and the reverse is true of business analysts and operators. High Technical skill ceiling Systems-first Translator (rare) Business-first Platform / Infra Engineer warehouse & orchestration only **Platform / Infra Engineer**Owns warehouse infrastructure, orchestration (Airflow etc.), and access controls: pure infra, no modeling or analysis. Usually splits out of Data Engineer once pipeline count or compliance needs grow. ML Engineer serving, deployment **ML Engineer**Productionizes the models a Data Scientist builds: deployment, monitoring, serving infrastructure. Rarely builds the model itself; a software engineer by training more than a statistician. Data Engineer pipelines, reliability **Data Engineer**Builds and maintains the pipelines and warehousing that make data reliable, available, and queryable. Optimizes for cleanliness, correctness, latency, and cost. The root of the systems chain. Data Scientist stats, experiments, readouts **Data Scientist**Statistical modeling, experimentation (A/B tests), ML, and forecasting. Heavier on math/stats and Python/R, lighter on dashboarding. The root of the applied-ML chain. Unicorn This person does exist, but they work at a frontier lab and you can't afford them. **Unicorn**Full stack: infra, modeling, stats, and the business fluency to explain all of it to the board. If you find one who isn't already at a frontier lab, hire them on the spot and don't ask about budget. Analytics Engineer semantic & metrics layer **Analytics Engineer**Owns the transform layer (dbt models, semantic layer, metrics definitions) so analysts work from clean, tested, documented tables instead of raw sources. Often confused with Data Analyst once the title lags the actual work. Data Governance docs, lineage, access **Data Governance**Owns documentation, lineage, access controls, and PII handling. Sometimes sits under infosec or IT rather than the data org, especially once compliance requirements show up. BI Engineer dashboard semantics **BI Engineer**Sits between Analytics Engineer and Analyst: owns the BI tool layer (Looker, Tableau) and the semantic modeling specific to dashboards. A further split of Analytics Engineer's job once it outgrows one person. Data / Analytics Manager roadmap, people **Data / Analytics Manager**People management plus roadmap, usually a former IC in one of the other roles. Not a distinct skill set of its own; that's why it sits outside every chain on this chart. Data Analyst SQL, dashboards, ad hoc **Data Analyst**Answers specific business questions, builds dashboards and reports, does ad hoc SQL analysis. Lighter on infra, heavier on business context and communication. The root of the analysis chain. Business Analyst process, requirements **Business Analyst**Less technical than a Data Analyst but with deeper subject-matter expertise: process analysis, requirements gathering, translating business needs into specs. Often collapses into Data Analyst at smaller companies. Big 4 Consultant accountability sink **Big 4 Consultant**Low technical and domain expertise. Somehow more expensive than the unicorn. Low Business / domain expertise High Systems chain Data Engineer → Platform/Infra, Governance, Analytics Eng. → BI Engineer Analysis chain Data Analyst → Business Analyst Applied ML chain Data Scientist → ML Engineer Orthogonal Data/Analytics Manager (not part of a chain) The darkest dot in each chain is the role that absorbs the others when headcount is tight; and the lightest is the role that gets folded back in first. Data Scientist and Analytics Engineer are commonly conflated in job postings; same for Data Engineer and Data Analyst. Both axes in the chart above are skills that take time and experience to develop, and therefore are somewhat mutually exclusive. Technical depth and business/domain expertise are both things people build with exposure and effort. A Platform Engineer can learn enough of the business to write a useful dbt model, a Business Analyst can learn enough SQL to pull their own numbers. A brilliant Data Engineer isn’t fluent in the business just because they’re fluent in the warehouse, and a sharp Business Analyst can’t write production pipelines just because they know the P&L. The key to hiring data roles correctly is being able to reconcile the requirements of the business from a data perspective with the competencies built by candidates in each of these roles. Below we've added a tool that lets you play around with various configurations of the 10 roles. Select any of the tiles individually or together, or use one of the presets we've added to see how they overlap. ## All ten roles, side by side Systems chain Analysis chain Applied ML chain Orthogonal (manager) Click any role to add or remove it from the comparison chart below. Data Engineer Platform / Infra Engineer Analytics Engineer Data Governance BI Engineer Data Scientist ML Engineer Data Analyst Business Analyst Data / Analytics Manager Data/Analytics Manager renders as a flat, undifferentiated shape on purpose: none of these twelve axes is "people management," which is most of the actual job. ## Compare any roles Pick two or more roles above and they'll stack here, or start from a preset team shape: Click roles in the grid above to compare them here. ## **Where AI-era roles fold in** In the AI era, these roles are further complicated by the addition of new roles that are adjacent to existing data roles. In many cases the AI role is really the existing data role wearing a moustache, (i.e. a Data Platform Engineer and a AI Platform Engineer use basically the same processes across different substrates), but for others, the jury is still out on whether they'll become their own role, or get folded into an existing one. The chart below shows the AI roles and their relationships in addition to the existing data roles, hover over any dot to see a description of the role and its overlap. High Technical skill ceiling Systems-first Translator (rare) Business-first Platform / Infra Engineer warehouse & orchestration only **Platform / Infra Engineer**Owns warehouse infrastructure, orchestration (Airflow etc.), and access controls: pure infra, no modeling or analysis. Usually splits out of Data Engineer once pipeline count or compliance needs grow. ML Engineer serving, deployment **ML Engineer**Productionizes the models a Data Scientist builds: deployment, monitoring, serving infrastructure. Rarely builds the model itself; a software engineer by training more than a statistician. Data Engineer pipelines, reliability **Data Engineer**Builds and maintains the pipelines and warehousing that make data reliable, available, and queryable. Optimizes for cleanliness, correctness, latency, and cost. The root of the systems chain. Data Scientist stats, experiments, readouts **Data Scientist**Statistical modeling, experimentation (A/B tests), ML, and forecasting. Heavier on math/stats and Python/R, lighter on dashboarding. The root of the applied-ML chain. Unicorn This person does exist, but they work at a frontier lab and you can't afford them. **Unicorn**Full stack: infra, modeling, stats, and the business fluency to explain all of it to the board. If you find one who isn't already at a frontier lab, hire them on the spot and don't ask about budget. Analytics Engineer semantic & metrics layer **Analytics Engineer**Owns the transform layer (dbt models, semantic layer, metrics definitions) so analysts work from clean, tested, documented tables instead of raw sources. Often confused with Data Analyst once the title lags the actual work. Data Governance docs, lineage, access **Data Governance**Owns documentation, lineage, access controls, and PII handling. Sometimes sits under infosec or IT rather than the data org, especially once compliance requirements show up. BI Engineer dashboard semantics **BI Engineer**Sits between Analytics Engineer and Analyst: owns the BI tool layer (Looker, Tableau) and the semantic modeling specific to dashboards. A further split of Analytics Engineer's job once it outgrows one person. Data / Analytics Manager roadmap, people **Data / Analytics Manager**People management plus roadmap, usually a former IC in one of the other roles. Not a distinct skill set of its own; that's why it sits outside every chain on this chart. Data Analyst SQL, dashboards, ad hoc **Data Analyst**Answers specific business questions, builds dashboards and reports, does ad hoc SQL analysis. Lighter on infra, heavier on business context and communication. The root of the analysis chain. Business Analyst process, requirements **Business Analyst**Less technical than a Data Analyst but with deeper subject-matter expertise: process analysis, requirements gathering, translating business needs into specs. Often collapses into Data Analyst at smaller companies. AI/ML Platform Engineer folds into Platform/Infra Eng. **AI/ML Platform Engineer**Runs the infrastructure for serving LLMs in production: model serving, vector databases, retrieval pipelines, inference cost and latency. Folds into Platform / Infra Engineer — same discipline, new substrate: warehouse infra → inference infra. Agent / Workflow Orchestration Engineer folds into Platform/Infra Eng. **Agent / Workflow Orchestration Engineer**Designs multi-agent systems and automation pipelines: the same fan-out, retry, and state-management discipline as a batch DAG, aimed at agents instead of jobs. Folds into Platform / Infra Engineer — same discipline, new substrate: Airflow DAGs → agent graphs. Data / Knowledge Curator for AI folds into Data Engineer **Data / Knowledge Curator for AI**Curates and cleans corpora for training, fine-tuning, and RAG: dedup, licensing, chunking strategy. A different quality bar than warehouse tables, same underlying discipline. Folds into Data Engineer — DE already owns vector DB infra at most companies; this is the content-quality half of that job. AI Governance / Responsible AI Analyst folds into Data Governance **AI Governance / Responsible AI Analyst**Model risk, bias/fairness review, and regulatory compliance (EU AI Act and similar) for AI systems specifically. Folds into Data Governance — direct counterpart: same function, model risk instead of PII/access risk. Eval Engineer folds into Data Scientist (stretch) **Eval Engineer**Builds eval harnesses, golden datasets, and scoring rubrics to measure whether a model, prompt, or agent change actually made things better. Folds into Data Scientist, but it's a stretch — borrows DS's experimentation rigor, but eval volume tends to outgrow a part-time host once model changes ship constantly. AI Engineer folds into ML Engineer (stretch) **AI Engineer**Builds products on top of foundation models (RAG pipelines, agents, tool use) without training models from scratch. Sits between software engineer and ML engineer. Folds into ML Engineer, but it's a stretch — the weakest fold-in here; pulls as much from general software engineering as from the data org, so it's often a net-new hire, not an absorbed responsibility. Low Business / domain expertise High Systems chain Data Engineer → Platform/Infra, Governance, Analytics Eng. → BI Engineer Analysis chain Data Analyst → Business Analyst Applied ML chain Data Scientist → ML Engineer Orthogonal Data/Analytics Manager (not part of a chain) AI-era roles Lighter tint, same color as the classic role they fold into (dashed line shows the fold-in target) ## **Why these roles are so different from one another** There are two reasons these roles are so heavily differentiated: The first is that data roles developed over the past 2 decades following a relatively non-traditional path. For this reason, many data professionals either migrated left or right in their organizations to end up in data; starting as a business analyst or statistician and moving left up the stack or starting as a software engineer and moving right. As a result, data professionals come from diverse backgrounds, and therefore can have wildly different core competencies. The second reason is that until recently, there was no classical training at the secondary or continuing education levels for data disciplines specifically, and standards and norms within the industry change extremely rapidly with subsequent advancements in technology. This contributes additionally to the spread of backgrounds and skills across data professionals. ## **Cheat Sheet** | Core role | Derivative roles | AI roles | |----------------|-----------------------------------------------------------------------------|-------------------------------------------------------------------------------------------------------------------------------------------| | Data Engineer | Platform / Infra Engineer, Analytics Engineer, Data Governance, BI Engineer | AI/ML Platform Engineer, Agent / Workflow Orchestration Engineer, Data / Knowledge Curator for AI, AI Governance / Responsible AI Analyst | | Data Analyst | Business Analyst | — | | Data Scientist | — | Eval Engineer | | ML Engineer | — | AI Engineer | [Kevin Koenitzer](https://www.linkedin.com/in/kevin-koenitzer-b00013a8/) A co-founder of Snowpack Data, Kevin has built ML applications, databases, and led analytics teams across industries. He excels at co-creating data strategies with his clients, providing product vision and clear direction. ## More from the blog [October 27, 2026 **Duck on the Edge** \ A pattern for operational analytics in 1st party product](https://snowpack-data.com/blog/duck-on-the-edge) [September 3, 2026 **Two Signals That Decide When to Build MLOps** \ A framework for investing in MLOps before it becomes an operational burden](https://snowpack-data.com/blog/two-signals-that-decide-when-to-build-mlops) [May 20, 2026 **The Context Your Analytics Agent is Missing** \ Why agentic analytics makes the case for a metrics layer stronger than ever.](https://snowpack-data.com/blog/context-your-analytics-agent-is-missing) ## References