STATUS: Open to relocation — UAEIndian NationalHyderabad, IN 500076Available: Immediate
Data Scientist — GenAI · LLM · Machine Learning
Jyothi S
Building production RAG pipelines & ML systems, 4.8+ years in
Data Scientist specializing in Retrieval-Augmented Generation, LLM training-data design and prompt engineering, with hands-on GCP data pipelines. Delivered a 100% on-time record across 6 concurrent AI initiatives for a multinational client (Google) with zero escalations.
Building and deploying ML & generative AI systems in production, across Epsilon and Indium.
tap for detail
100%
On-time delivery
Across 6 concurrent AI initiatives for Google, zero escalations for accuracy or delay.
tap for detail
30%
Model accuracy ↑
Tuning classification & regression models with Python and Scikit-learn at Epsilon.
tap for detail
25%
Processing time ↓
Redesigning and automating ML preprocessing pipelines at Epsilon.
tap for detail
Skill retrieval consolequery → type or click a skill
Filters the experience below by relevance — the same idea as the reranking layer she built for retrieval matching. Skill terms in the matrix further down are clickable too.
Experience — click a role to expand/collapse
Data Scientist · Indium
Jan 2025 – Present · Hyderabad, India ▼
Built and deployed an end-to-end RAG system for a Grocery Chatbot using LangChain, sourcing product data from Uber Eats' internal database as the retrieval knowledge base for natural-language product search.
Designed an LLM-based reranking layer on top of the retrieval pipeline to re-score and reorder candidate matches, directly improving exact-match accuracy over baseline similarity search.
Designed and built the end-to-end Connector Hub data platform for Google AgentSpace — scalable data-ingestion and dataset-creation pipelines across Drive, Calendar, email, Jira, ServiceNow, Notion and Canvas using Python, SQL, Gemini CLI and GCP.
Built data quality frameworks (schemas, labeling rules, QC checks) for ML and LLM evaluation datasets (query-answer pairs), improving model benchmarking accuracy and coverage for AI agent development.
Delivered prompt engineering and LLM training data design across 6 concurrent generative AI initiatives for a multinational client (Google), maintaining a 100% on-time delivery record with zero escalations.
Built SQL-based monitoring and validation logic for ML/OCR pipelines (Menu Reader, Receipt Digitization) to track model error rates and SLA metrics, partnering with engineering on fixes.
Designed a rule-based risk-scoring model for PRDs via SQL pipelines over UPlan analytics tables, enabling stakeholders to prioritize the highest-risk documents for intervention.
ML Developer · Epsilon
Sep 2021 – Jan 2025 · Bengaluru, India ▼
Increased marketing model accuracy by 30% by building and tuning classification and regression machine learning models using Python and Scikit-learn.
Reduced data processing time by 25% by redesigning ML workflows and automating data preprocessing pipelines, improving resource utilization.
Uncovered actionable customer trends, behaviors and outliers through exploratory data analysis across multi-source marketing datasets, directly informing targeted marketing strategy.
Improved data integrity and consistency across analytics pipelines using SQL to extract, clean and preprocess large-scale datasets from diverse sources.