Senior Director, RGM Product AI Engineering

The Coca-Cola Company•Atlanta, GA

4d•$202,000 - $229,000•Onsite

About The Position

Revenue Growth Management (RGM) is a critical commercial capability at The Coca-Cola Company, focused on driving topline growth by connecting consumer demand to system economics. RGM determines product offerings, packaging, pricing, channel strategy, and promotional investments to achieve balanced growth across transactions, volume, and price/mix. The company is currently transforming RGM into an intelligent decision platform for the Agentic AI era, where AI agents will interact with enterprise data, utilize advanced analytics and optimization models, run scenarios, and provide recommendations with justifications to commercial teams. This initiative is large-scale and high-impact, involving agent orchestration, retrieval over enterprise data, and the evaluation of complex systems. The work is not theoretical; it is intended for immediate use by commercial teams to make real-time decisions. This role offers the opportunity to build the next generation of commercial decision intelligence at Coca-Cola and establish the technical foundation for agentic RGM at a global scale. The patterns developed here will influence how one of the world's largest commercial systems makes its most critical decisions. The position is designed for individuals who want their work to be utilized by real users at scale from the initial release.

Requirements

Bachelor’s degree in Computer Science, Engineering, Data Science, Information Technology, or a related technical field; Master’s degree preferred.
8+ years of software, data, or ML engineering experience, including deep hands-on Python running in production and operated by other people — services or pipelines rather than notebooks alone — with strong pandas and SQL.
Demonstrated experience shipping an LLM-enabled feature that real users used, with a clear account of what broke first and how it was caught.
Treats prompts as code: versioned, tested against a fixed input set, with one variable changed at a time.
Experience building LLM-enabled systems, including retrieval-augmented generation, text-to-SQL, prompt and agent orchestration frameworks (e.g., LangGraph, LangChain), vector stores, tool/function calling, agentic workflows, and human-in-the-loop guardrails.
Working knowledge of evaluation practice for generative systems: fixed test sets, rubric-based grading, hallucination and coverage measurement, and eval-gated release.
Ability to read price, volume, and revenue data by SKU and channel, and to spot claims the numbers do not support.
Precise written English; the quality of the pipeline’s output is bounded by the quality of its instructions.
Understanding of how to integrate AI applications with traditional machine learning, analytical, and optimization models.
Strong understanding of modern software engineering practices, including automated testing, CI/CD, source control, observability, API design, security, and cloud deployment.
Working knowledge of the Azure data and AI stack, including Azure Data Lake Storage, Databricks, Synapse, Azure Data Factory, MLflow, and Azure OpenAI Service, plus modern application patterns such as API-first middleware (FastAPI, Azure App Service, Azure API Management), Azure SQL, Key Vault, Azure Active Directory/Entra ID, Azure Monitor, and Azure DevOps CI/CD.
Strong communication skills, with the ability to explain model behavior, its limits, and its evidence to commercial audiences in business-relevant language.
Comfort operating in a matrixed, multi-market franchise environment where adoption depends on trust in the output.
A global role supporting operating units and bottlers in every region requires working across time zones, including early and late calls and availability outside standard business hours when markets or releases de

Nice To Haves

Familiarity with Revenue Growth Management concepts across pricing, promotion, assortment, and mix, and with the analytics behind them — elasticity modeling, optimization algorithms, and forecasting; preferred.
Exposure to CPG or bottler commercial data, including how pricing and promotion decisions cascade to execution at the point of sale; preferred.
Sufficient C# to read and reason about the .NET service layer the pipeline integrates with; preferred.

Responsibilities

Own the prompt behind each pipeline node — validation, narration, slide planning, and QA review — versioned in source control alongside test inputs and expected outputs rather than maintained in a chat window.
Extend the LangGraph orchestration pipeline in Python: ingest run exports, compute KPIs with pandas, call the model for narrative, and route QA retries when a generated slide fails review.
Define and track the evaluation metrics for every run — the share of insights analysts rate gold or silver, the share hallucinated or missing a baseline, and volume coverage — and gate prompt changes on those numbers improving.
Build retrieval over reference material, including methodology documentation, QA references, and past runs, so generated text quotes the underlying data instead of guessing.
Prototype new analyses on existing run data, from new insight types to competitor-response summaries, promoting a prototype to a feature only when it has a test set and a user who asked for it.
Integrate the pipeline with the platform alongside the Tech Lead, ensuring every run artifact — prompts, computed KPIs, narrative, and QA verdicts — lands in the SQL Server schema with logging and observability.
Design human-in-the-loop guardrails so commercial users can see, challenge, and correct what the system generates before it reaches an OU or bottler audience.
Make and document technical decisions on model providers, orchestration approach, and hosting within enterprise security, privacy, and AI governance requirements.
Integrate LLM-based capabilities with the traditional machine learning, analytical, and optimization models that produce the underlying pricing and promotion recommendations.
Design for the system context: multi-market, multi-OU deployment, bottler-facing data boundaries, and role-based access to commercially sensitive pricing and promotion data.
Partner with the business consultant on insight quality, turning recurring user feedback into concrete pipeline and prompt changes and retesting with the same users once a fix ships.
Contribute to engineering standards for a codebase with non-deterministic components: testing, observability, and eval-gated releases.