Lead Agentic Data Systems Engineer
Salesforce.com, inc. · Ciudad de México, Mexico ·
- Seniority
- Lead
- Employment
- Full time
- Category
- Software engineering
- Experience
- 5+ years
Salesforce.com, inc. · Ciudad de México, Mexico ·
Mexico City | Full-Time
Hybrid
Department Overview
The Enterprise Data & AI Solutions group is the organization’s strategic hub for cognitive automation. We move beyond traditional data management to build the autonomous engines that power executive decision-making. Our team is composed of Architects of Autonomy—professionals with the technical depth to build systems from the ground up and the strategic vision to leverage AI to ensure scalability. We partner with the C-suite to solve high-complexity challenges by deploying sophisticated multi-agent ecosystems that operate with continuous uptime.
We are seeking a Lead Agentic Data Systems Engineer.
This is a role for a hands-on, depth-first engineer who will take the architectural blueprints set by our Principal Engineers and turn them into hardened, production-grade data products. This role is defined by execution depth. You will own the product end-to-end — building it, maintaining it, enhancing it, and constructively challenging the design when implementation reality demands it. You are the person the business trusts to make a data product actually work, at quality, every day. You are someone who knows how to supercharge their own workflow with AI agents, but your primary leverage comes from deep, disciplined building rather than broad orchestration.
The Strategic Shift: You are redefining the data team model. Instead of managing human personnel, you manage complex "hand-off" protocols between specialized AI agents, acting as the central anchor for a hybrid human-agent intelligence unit.
Autonomous Scaling: Architect and maintain a private ecosystem of 10+ autonomous agents specialized in ETL, synthetic data generation, automated QA, and predictive modeling.
Agentic Orchestration: Design multi-step reasoning architectures and verification protocols to ensure agents autonomously validate and peer-review their own outputs.
Complex Problem Resolution: Transform high-level, ambiguous business requirements into production-ready data products independently, bypassing the need for mid-level project management.
Governance & Oversight: Use domain knowledge to ensure deployed tools are well governed. Governance as code for data pipelines and Agentic development. Context aware Agent development.
Contextual Integration: Develop and maintain Model Context Protocol (MCP) servers to provide agents with secure, deep-link access to Snowflake, Salesforce, AWS, and proprietary internal data catalogs.
Engineering Foundation: Production-grade proficiency in Python, dbt, Airflow, and advanced SQL. Apache Spark, and Snowflake.
AI Orchestration: Fluency in AI-native development environments (e.g., Cursor, Codex, or Claude Code). Expert in Prompt Engineering. Mastery of agentic frameworks such as LangGraph. Leverage MCP servers to retrieve data from tool stack
Cognitive Architecture: Expert-level knowledge of chain-of-thought prompting, self-correction loops, and iterative reasoning paths.
Salesforce Knowledge: Salesforce Core and Data 360 understanding
Systems Design: Advanced understanding of Data Mesh, Data-as-a-Product (DaaP), and Event-Driven Architectures. Semantic layer. Knowledge Graphs.
Cloud Infrastructure: Experience using agentic workloads via Docker, Kubernetes, and serverless compute environments.
Experience: 5+ years of experience in high-stakes Data Engineering, Architecture, or Data Science.
Strong Python / SQL Expertise
Operational Leverage: A documented history of using generative AI to accelerate personal and departmental output by orders of magnitude.
Strategic Autonomy: The ability to function as a "Domain Data Officer," managing end-to-end data strategy for a business unit with minimal supervision.
Technical Intuition: Superior analytical judgment—the ability to identify subtle logic errors or hallucinations in agentic output before they reach production.
08:00 – Intelligence Synthesis
While you start your day, your "Scout Agents" have already completed an automated audit of the overnight data pipelines. You review a synthesized report highlighting three anomalies in the global revenue stream. One agent has already drafted a proposed SQL remediation and a unit test; you review the logic and authorize the deployment to production.
Stakeholder triage and problem framing/overnight pipeline audit, anomalies, drafted SQL remediation, unit test, approve deployment:
Claude Code / Claude Enterprise / Codex / Agentforce-style command layer
Snowflake, Tableau metadata, dbt artifacts, pricing event logs
ranked brief with pricing overrides, lineage breaks, semantic definitions, supporting evidence
human deciding noise vs escalation vs today’s work
10:30 – Architectural Orchestration
A request arrives from the CFO for a "High-Resolution Market Volatility Stress Test." Rather than building the model manually, you define the parameters for your agentic fleet. You orchestrate a "Research Agent" to pull external market indicators, a "Simulation Agent" to run the Monte Carlo iterations, and a "Synthesis Agent" to build a live-updating executive dashboard. You spend your time on validation and strategic interpretation.
13:30 – Knowledge Retrieval & Documentation
You encounter a legacy pricing algorithm with no surviving documentation. You deploy an "Information Retrieval Agent" to parse thousands of historical Slack threads, Jira tickets, and GitHub commits. Within minutes, the agent provides a technical summary of the original design intent. You direct the agent to update the global metadata repository so this knowledge is permanently accessible to the organization.
legacy algorithm overview/chage, retrieval across historical Slack / Jira / GitHub, technical summary, metadata repository update.
leadership-triggered critical investigation
search across SQL corpus, dbt lineage, notebook agents
drafted explanations and SQL patches
converting findings into a leadership action plan
15:30 – Defensive Systems Engineering
You dedicate time to "Security & Integrity Engineering," building new "Red-Team Agents" whose sole purpose is to attempt to find flaws in the logic or security vulnerabilities in your other agents. You are building a self-healing digital immune system for the company's data.
Agents prepare modeling and simulation workflows to test strategic scenarios such as pricing or renewal changes. The human selects methods, reviews assumptions, and interprets uncertainty before any decision is actioned.
18:00 – Asynchronous Task Deployment
You initialize a long-tail analytical task: "Analyze the last 24 months of customer churn data and identify latent correlations that current linear models have missed." You disconnect while the agentic fleet begins the heavy compute and reasoning cycles overnight.
20:00+ – Long-Tail Agent Execution After hours, agents continue computationally heavy or long-horizon tasks, creating a curated queue of opportunities, risks, and partially completed work for the next morning.