Applied Data Scientist, Cresta

• 3+ years building and shipping models for real-world business applications, ideally in NLP and LLM-based systems
• Strong proficiency in Python and standard ML / data tooling (e.g., SQL, data pipelines, experiment frameworks)
• World-class first principles thinking and ML intuition
• Ability to turn ambiguous product asks into crisp problem statements, eval specs, metrics, and hypotheses
• Experience working directly with customers or internal stakeholders to understand constraints, explain tradeoffs, and iterate on solutions
• Comfort working with design-partner style engagements where requirements evolve rapidly and you’re expected to co-create the solution
• Track record of building evaluation suites that go beyond single scalar metrics to capture reliability, safety, and qualitative user experience
• Strong written and verbal communication skills; able to clearly explain complex technical work to both engineers and non-technical partners

Desirable

• As an Applied Data Scientist on our Insights team, you will help pioneer the next generation of Cresta features that power how enterprises understand and optimize every customer interaction
• You’ll partner directly with Product Managers, Machine Learning Engineers, other Data Scientists, and a set of strategic design-partner customers to solve their biggest challenges
• If you love working from first principles in a rapid iteration loop with customers - this is your chance to have an outsized impact on the customer experience industry
• Co-develop new capabilities with a small number of high-impact enterprise customers along with our product, engineering, and design teams; using their real workflows and constraints as your testbed
• Communicate effectively across all levels of the organization
• Plan and run short, focused design-partner engagements (days to weeks) where you ship early versions, collect structured feedback, and iterate quickly
• Generalize learnings from each design partner into reusable, productized capabilities rather than one-off bespoke models
• Partner with domain experts to curate high-quality eval guidelines and datasets for domains such as CSAT prediction and outcome prediction (across both human<>human and human<>AI interactions)
• Use the best tools + models for the job (simple and interpretable where it matters, sophisticated where it can drive outsized value
• Write clear specs and experiment reports that make tradeoffs and assumptions explicit
• Stay close to the research frontier in ML/AI, LLMs, and evals, translating promising ideas into pragmatic, shippable improvements
• Where applicable, help translate your solutions into publications, whitepapers, technical blogs, etc