What she does best
Designs AI products at the intersection of machine learning, document intelligence, cloud platforms, and operational scale.
Data Scientist • Generative AI • Agentic Systems
Gayathri Kumar builds intelligent systems that turn complex workflows into scalable products.
Dallas-based AI engineer with 4+ years in Generative AI and 7+ years of end-to-end delivery across Deloitte, Ford Motors, Bosch & Siemens, and RSA Canada.
Profile
Gayathri Kumar builds AI systems that turn complex workflows into measurable impact.
Where she stands out
Turns ambiguous business workflows into measurable automation with RAG pipelines, computer vision, predictive models, and robust data systems.
How she works
Combines technical depth, cross-functional collaboration, and production-minded delivery to move AI from concept to reliable business value.
Experience
Work that blends research-minded AI with enterprise execution.
Cloud Data Transformation & Analytics Automation
- Migrated 5M+ records into a scalable Azure cloud ecosystem with full governance compliance.
- Automated ETL workflows with Azure Data Factory and Databricks, accelerating delivery by 3–4 weeks.
- Built Tableau and Plotly dashboards that increased executive decision speed by 39%.
- Received a performance-based spot award for technical excellence and leadership.
Legacy Engineering Workflow Automation
- Replaced manual blueprint data entry with an AI extraction ecosystem, recovering 80+ hours per week.
- Built OCR and computer vision pipelines with OpenCV and Tesseract for technical drawing intelligence.
- Improved data accuracy by 15% and aligned automation with automotive safety compliance standards.
AI-Powered Facial Analytics & Underwriting Intelligence
- Built a transfer-learning classification pipeline on a self-labeled 5,000+ image dataset.
- Reached 88% accuracy in tobacco-induced aging biomarker detection.
- Automated landmark detection and normalization for scalable underwriting and fraud detection.
Operations Automation & Reporting
- Architected web-based automation tools that reduced product operations latency by 30%.
- Built leadership dashboards that eliminated around 10 hours of manual reporting per week.
Insurance Systems Quality & Reliability
- Resolved a decade-old production defect through deep data analysis and received formal client commendation.
- Orchestrated end-to-end QA lifecycles and automated regression suites for complex insurance workflows.
Selected Projects
Interactive snapshots of applied AI work.
Agentic AI
AI-Powered Document Intelligence & Cross-Reference System
Built a smart AI system that searches and cross-references thousands of engineering documents and blueprints, eliminating hours of manual review.
- Added automated quality checks to validate AI output accuracy.
- Used MCP, ReAct, multi-agent orchestration, VLMs, G-Eval, KV caching, and ChromaDB.
LLM / RAG
LLM Fine-Tuning, Alignment & RAG Pipeline
Fine-tuned a large language model for more accurate conversation summarization and aligned it with human feedback for safer, more reliable responses.
- Built a context-aware retrieval pipeline for large unstructured document libraries.
- Used FLAN-T5, LoRA, PEFT, RLHF, PPO, LangChain, watsonx.ai, and ROUGE metrics.
Computer Vision
Facial Analytics for Underwriting Intelligence
Built a facial analytics pipeline for biomarker detection using transfer learning and automated landmark normalization.
- Trained on a self-labeled 5,000+ image dataset.
- Reached 88% classification accuracy and improved underwriting scalability.
Skills
A modern AI stack spanning orchestration, modeling, and cloud delivery.
Generative & Agentic AI
MCP, LangGraph, CrewAI, AutoGen, ReAct, multi-agent orchestration, RLHF, GRPO, self-correction loops.
RAG & LLM Engineering
Multimodal RAG, GraphRAG, semantic chunking, VLMs, prompt engineering, PEFT, LoRA, fine-tuning, LLM evaluation.
Data & Cloud
Azure Data Factory, Databricks, AWS, PySpark, ETL orchestration, Docker, Kubernetes, Medallion architecture.
ML, NLP & Vision
Transformers, TensorFlow, PyTorch, OpenCV, Tesseract, predictive modeling, statistical analysis, reinforcement learning.
Languages & Analytics
Python, SQL, R, Tableau, Power BI, Plotly, Matplotlib, Pandas, NumPy, Scikit-learn, ChromaDB.
Working Style
Cross-functional delivery, enterprise modernization, compliance-aware automation, measurable business impact.
Education
Academic grounding paired with applied delivery.
Master of Science in Information Science
University of North Texas, Denton, Texas
Concentration in Information Systems • GPA 4.0/4.0 • Aug 2024 – May 2026
Contact
Let’s build practical AI systems that feel ambitious and reliable.
Available for data science, AI engineering, and applied Generative AI roles focused on production-grade systems.