Data Scientist III

Job Role

As a Data Scientist III, you will own the analytical strategy for a core product area and take it all the way to a governed, monitored production system at scale, not just a working notebook. This role owns model-serving workflows end-to-end: you will define and optimize the data pipelines and ETLs that feed your models, build models that hold up under real production load, and implement the monitoring and governance that keeps them trustworthy over time. You will also mentor other data scientists doing the same.

Key Responsibilities

Own the end-to-end analytical framework for a product area, i.e., from problem framing through production deployment and monitoring, with minimal oversightApply LLMs, prompt engineering, and retrieval-augmented generation (RAG) to real product problems — e.g., translating complex targeting/attribution outputs into plain language insights, or building AI-assisted workflows for internal or client-facing tools.Evaluate when generative AI is the right tool versus when a simpler model or heuristic will outperform it. This role is expected to have informed judgment here, not default-to-LLM instinctsStay hands-on: personally build, code, and develop AI/ML solutions rather than only architecting and delegatingBuild models that scale to production data volumes and real-time serving loads, not just prototypes that work on a sampleDefine and optimize data pipelines, ETLs, and AI model-serving workflows for your domain; this is not handed off to a separate MLOps functionImplement model monitoring and governance (drift detection, performance tracking, retraining triggers, audit trails) so that production models stay reliable and defensible under compliance scrutinyDesign measurement approaches for problems where clean ground truth doesn't existMentor and provide technical guidance to Data Scientist I/II team members; review and approve their modeling and code before it ships.Design and defend experimentation approaches (A/B testing, causal inference, uplift modeling) that hold up under stakeholder and compliance scrutinyDrive adoption of better MLOps tooling, scalability practices, or modeling methods across the data science teamProactively identify whitespace opportunities in product roadmap and bring data-backed recommendations to stakeholders, not just answers to their questions.

Skills Required

Education: B.Tech / MS in Computer Science, Data Science, or a related field.At least 5 years of hands-on experience in data science and machine learning (both) is a must haveHands-on MLOps experience is a must have. You should be comfortable owning a model from training through deployment, serving, and production monitoring, not handing it off once it works in a notebook. This includes experience with model-serving infrastructure, pipeline/ETL orchestration, and monitoring or governance tooling (drift detection, performance dashboards, retraining triggers)Proven experience building and scaling machine learning models to handle production data volumes, throughput, and latency requirements is a must have. This means designing for scalability up front, not retrofitting it after something breaks in productionHands-on experience with LLMs and RAG architectures is a must have. This includes practical experience with prompt engineering, embedding-based retrieval, and vector databases (e.g., Pinecone, FAISS, Weaviate), and frameworks like LangChain/LangGraph or AWS BedrockDemonstrated ownership of at least one modeling project taken fully end-to-end into production, with measurable business impactExperience designing and running experiments (A/B testing, or causal inference methods such as DiD, IV, PSM, uplift modeling) is non-negotiableTrack record of mentoring or reviewing the work of junior data scientists is a must haveA strong knowledge of SQL, data structures, and query optimization at a level where you can review others' queries, not just write your ownExperience fine-tuning generative AI models (not just calling APIs) for real-world applications is highly desirableFamiliarity with AI agent orchestration and multi-step agentic workflows is a plusPrior experience in advertising technology, programmatic media buying, or another high-scrutiny/regulated data domain is highly advantageous and good to have.

Personality & Work Ethic

A strategic thinker who sets technical direction for a problem area, not just aligns to oneComfortable operating in ambiguity where the "right" ground truth doesn't exist yetExcellent communication skills, with a track record of influencing product or technical decisions using dataGenuine mentorship instinct; has grown at least one junior team member's skills, not just supervised their outputOwnership mentality that extends past "the model works" to "the model is scalable, monitored, governed, and still trustworthy six months from now."Doesn't let a research rabbit hole delay a hard deadline

Skills: etl pipelines,machine learning,faiss,llm,model training,data modelling,pinecone,langchain,mlops,a/b testing,data science,langgraph