Sr. Data Scientist - Agentic AI

Job Title: Sr Data Scientist, Agentic AI Bengaluru,India “ This is a hybrid position, which will require the ability to be onsite in our Bengaluru, India office as needed. Candidates must be based in India. ”  ABOUT THE COMPANY: Clari + Salesloft are building the next era of enterprise revenue — one where teams make confident decisions powered by AI and real signals. By combining our scale, insights, and AI innovation, we’re building the industry’s first Predictive Revenue System , enabling humans and AI to work together to make smarter decisions and drive consistent growth. With thousands of customers using our platforms every day, we have an unmatched view into how revenue is actually won — the Revenue Context that reveals what happens, when, and with what outcome. This gives us a unique opportunity to transform an entire category and set a new benchmark for how modern revenue teams operate. Join us to help transform how companies around the world run revenue — and build the platform that will guide leading revenue teams into the future. THE OPPORTUNITY: At Clari + Salesloft, our Sr Data Scientist will be pivotal to our company’s success. Our Sr Data Scientist will be a hands-on builder, someone who takes an ambiguous agentic AI problem, turns it into a working prototype, implements it, and partners with engineering to get it into production. This is an execution role: you will spend most of your time prototyping, building eval. On a day-to-day basis, you will: Prototype fast: Turn a rough problem statement ("can we auto-summarize this deal", "can an agent update this CRM field safely") into a working prototype - RAG pipeline, LLM-based Inferences, or an agent - within days, not weeks. Implement for real: Take your own prototypes from notebook to production-quality Python code, tested, typed, and reviewable. Ship with engineering: Partner with the platform engineers who own the harness and infra to get your work live. You don't need to design the harness, but you need to understand it well enough to build within it and debug when something breaks. Own eval for all AI work: Build and run the evaluation for what you ship - offline eval sets, LLM-as-judge checks, and tracking regressions before and after changes. Use the right tool for the job: Apply classical NLP/ML where it's the better fit, and LLM-based approaches where they're not. You should have real judgment from having done both. Learn and adapt fast: Pull in proof-of-concepts, tests, and external research when the path isn't obvious, and course-correct quickly rather than waiting for direction. In addition to working with amazing colleagues who exemplify our 'team over self' core value, you will have the opportunity to build impactful, revolutionary software that is changing the way revenue is predicted. WHAT WE’RE LOOKING FOR: We're looking for a builder, not a manager-in-waiting. Someone who has spent the last 2-3 years genuinely shipping LLM-based inference, RAG, or agent work (frameworks, sometimes multi-agent) after 2-3 years of solid core NLP-driven data science before that. If you'd rather be hands-on-keyboard turning ambiguous problems into working software, this is the career path for you. THE TEAM: Clari + Salesloft's goal is to build the world's first AI-driven Predictive Revenue System. The Applied AI Engineering team supplies the AI platform, the harness, memory, retrieval, and tooling layers, that our agents and models run on. This role sits alongside that platform team as an execution-focused builder: taking the platform's building blocks and turning them into shipped, evaluated agentic capabilities for revenue teams. If you're energized by fast prototyping, real production ownership of your own work, and close collaboration with the engineers who run the platform underneath you, we'd love to meet you. Vision: Fundamentally transform the way buyers and sellers drive repeatable outcomes Mission: Bring science to the art of sales  THE SKILL SET: Experience: 6-8 years total, including 2-3 years hands-on in agentic AI and 2-3 years in core NLP/DS before that. Agentic AI - Hands-On Build List: Real, personal build experience with: LLM-based inference and Q&A systems in production  RAG platforms — retrieval pipelines, chunking, tuning for relevance  Agents built with an existing framework (LangChain, LlamaIndex, DSPy, or an in-house harness) Evaluation - offline eval sets, LLM-as-judge, regression tracking for what you ship Guardrails - input/output validation on your own agent's actions Multi-agent systems, MCP-style tool registries, Agent-to-Agent (A2A) coordination  [good to have] Graph RAG / knowledge-graph-backed retrieval  [good to have] Core NLP & ML Foundations: 2-3 years working in classical NLP/DS - text classification, NER, embeddings, feature engineering, before moving into LLM-era work. You should be able to say why a classical approach beats an