Shipping reinforcement learning that matters: Life as a Forward-Deployed Data Scientist at Braze

At Braze, our Data Science team doesn’t just build models in a vacuum. We’re working shoulder to shoulder with global brands, helping them move past generic customer communications to power truly personalized experiences at scale, thanks to BrazeAI Decisioning Studio™.

Let’s explore how we use reinforcement learning (RL) to drive results, what the journey looks like for a Forward-Deployed Data Scientist in our team, and why our "open-source" team culture is what makes this real-world impact possible.

Braze data science: The quick essentials

• The goal: Moving the needle on incremental uplift—think higher customer lifetime value (CLV) or higher revenue from a use case.
• The tech: Machine learning, with a focus on reinforcement learning.
• The stack: Python, SQL, Spark.
• The role: Data pipelines, feature design, causal/experimental thinking, model tuning, monitoring, and clear communication. As a Forward-Deployed Data Scientist at Braze, you’re part Machine Learning Engineer, part Strategic Consultant.
• The team: We are a globally-distributed, English-speaking team that shares knowledge fast and supports each other along every step on the journey to deliver results.
• The impact: RL allows brands to move from manual rules and static A/B tests to a system that learns from real user behavior and aligns with business goals. We work across countries, industries (retail, financial services, food, energy, and more), and lifecycle stages (acquisition, nurture, upsell, cross-sell, winback).

What we actually deliver with reinforcement learning

Most marketing teams have many ideas to test and very limited time to explore them. Plus, they are overwhelmed by choices: What to send, when to send it, and who should receive it. Historically, companies used static A/B testing to find an answer, but that only gives you a snapshot in time and they are quite labor-intensive to perform. We use reinforcement learning to build systems that help them move from rules and static A/B tests to a system that learns from real outcomes and adapts.

When we set up an RL agent, we define:

• Actions: The "levers" we can pull (e.g. creative templates, channels, or discount level).
• Context: What we know about the user right now (e.g. attributes, behavior, preferences, lifecycle stage).
• Rewards: The goal—whether that’s a customer lifetime value, a purchase, revenue, or downstream value.
• Constraints: The guardrails (e.g. "only recommend the most aggressive offers to the highest churn rate").

We align the RL reward with what the business cares about most, not just short-term clicks. That may be conversion, purchases, or user engagement. The system learns over time and adapts as users and markets change.

The Forward-Deployed Data Scientist journey: From design to performance

Being "forward-deployed" means you aren't just handed a data set—you help build the blueprint. Our journey has three different stages: Design, implementation, and getting to performance. As a data scientist, you partner with an Engagement Manager and an AI Success Manager to oversee timeline and stakeholders, so that you can focus on the business problem and the machine learning solution.

1. The design phase

We start by learning the customer’s goals and current setup. We run working sessions with Marketing, Analytics, and Engineering to understand:

• What the customer sells and what they want to grow
• What the customer journey looks like today
• What data exists and how reliable it is.

We also agree on measurement—that is, what “success” means, how we track it, and key trade-offs.

You aren't on your own here. We share our draft use cases with the full Data Science team at Braze in open forums and knowledge-sharing sessions. We also have thorough use case and general knowledge documentation. We ask ourselves, “Does this look like a use case we ran before?” and “What worked and what failed?”. This lets us reuse patterns, avoid known traps, and pick strong starting configurations. Because the team is accessible and so collaborative, it is easy to reach out to a peer who solved something similar in another region or vertical.

2. The implementation phase

Design becomes a live integration. The Forward-Deployed Data Scientist turns business goals into data inputs, features, and model configuration. This phase needs strong statistical thinking. We judge early signals, separate noise from signal, and align model behavior with business rules so the system is safe and useful.

• The work: You’ll be helping to add customer data to the platform, engineering and validating features, defining rewards and attribution windows, configuring the ML setup for the use case, and running analysis to confirm the system behaves as expected.
• The vibe: It’s deeply collaborative. We ask other Braze teams for implementation tips and share code snippets and checks. We meet with the Product team to review feature options, flags, and roadmap details when we need a custom setting. If we need help on data flows, we can talk to internal engineering. The norm is open support: Anyone within Braze can ask for help, and people will always respond.

3. The iteration phase

After launch, the system learns from real user interactions. This is where impact (and fun!) happens and where iteration truly matters.

• Continuous improvement: We use experimental design to prove our impact. You’ll be reviewing performance and model preferences, running experiments and reading results with care, tuning configuration, adjusting features, reward definitions, and attribution windows, and proposing optimizations.
• Non-stop collaboration: We partner with Product, Engineering, and other Data Science teams to dig into product details. We run problem-solving sessions to debug edge cases, discuss anomalies, and plan next tests. Global coverage means someone is always available to review a question—fast. This keeps learning loops short and helps us get to strong, stable uplift.
• Ongoing delivery: We don’t simply get you started and declare the project done. We are constantly working behind the scenes to support our customers improving uplift over time through proactive intervention and strong partnership with our customers. We update use case design as the customer’s needs and priorities shift, which could be as small as a new creative template or as big as a complete use case rebuild. In many ways, we actually act as a strategic extension of our customers’ marketing and data teams.

Why you’ll like the team

We are a united, global team. We meet weekly to share what worked, what did not, and which patterns are worth reusing. We also run problem-solving sessions on live use cases.

• Low ego, high impact: People are always eager to lend a hand across teams—Data Science, Product, Engineering, and Customer Experience. It is normal to DM someone you do not know yet to ask for context or a quick review. Knowledge is open by default: docs, playbooks, examples, and dashboards are easy to find, and people are happy to walk you through them.
• Global brain: With teams across the world, someone is almost always around to help you talk through a problem, and help with it.

Plus, Braze has an award-winning workplace culture—from employee resources groups and clubs to regular social events and volunteering opportunities, you’ll find a community where you can thrive.

Is the Forward-Deployed Data Scientist role for you?

The best Forward-Deployed Data Scientists are people who love the "how" and the "why" just as much as the "what."

• The technical side: You’re comfortable building and refining data pipelines with partners, designing features that reflect user behavior, applying experimental and causal thinking, and fine tuning models.
• The human side: You enjoy learning all about customers and their success metrics, translating requirements into a model-ready set up, proposing use cases that are measurable and scalable, and sharing insights that help marketing teams act.

At Braze, data scientists connect machine learning to real outcomes. If you want to ship reinforcement learning models that impact millions of users, help customers move from fixed logic to learning systems, and do it with a global team of passionate and kind people that have your back, we’d love to meet you.

We are regularly hiring forward-deployed data scientists to help us shape the future of AI-driven customer engagement. Find out more about working at Braze and check out our open roles.