Nikita Kumari
|
I build AI that's useful, not just impressive.
About Me
I started out in biology, running experiments and staring at spreadsheets, and realized the data was the part I couldn't stop thinking about. That pulled me into a Master's in Data Science at ASU.
These days I like taking a project all the way: messy data in, a working product out. My fraud detection project is a good example. It covers the PySpark pipeline on Databricks, a model I had to debug for data leakage, and a live API you can try right now. I'm also an AI Fellow at Handshake AI, where I spend my time thinking about how LLMs fail and how to measure it.
I'm actively looking for full-time ML/AI roles. Have a messy dataset, a wild idea, or an open position? I'd love to hear about it. Say hi below.
Technical Arsenal
A collection of the tools and technologies I use to turn data into value.
Programming & Databases
Machine Learning & NLP
Cloud, Big Data & MLOps
Data Visualization & BI
Core Skills & Tools
Certifications
Featured Projects
A selection of my work. Filter by category.
Real-Time Credit Card Fraud Detection
End-to-end fraud detection on 284K transactions (0.17% fraud): PySpark ETL into Delta Lake on Databricks, a stacked Random Forest + XGBoost ensemble with SMOTE and leakage-safe training, tracked in MLflow. Deployed as a containerized FastAPI service on Render with real-time scoring, SHAP explanations and KS-test drift monitoring. 78% recall at 88% precision (PR-AUC 0.83) on held-out data.
Medical AI Diagnostic Agent
This project is an AI-powered agent designed to assist with medical differential diagnosis. It leverages a Retrieval-Augmented Generation (RAG) architecture to provide evidence-based suggestions from a curated medical knowledge base, running entirely on local, open-source models.
SentinelRAG: Privacy-Preserving Compliance System
Privacy-preserving RAG for OFAC sanctions screening: hybrid retrieval (BM25 + vector search with re-ranking), a RAG evaluation framework (RAGAS), and a Microsoft Presidio-based privacy layer targeting zero hallucination tolerance across 100+ test cases with PII fully masked.
Dynamic Promotion Personalization
An A/B test simulation where a contextual multi-armed bandit learns to personalize marketing offers, achieving a +132% lift in conversions over a static control group.
Biomedical Literature Search & Insight Engine
A system that ingests biomedical research papers from PubMed, processes the text, and allows users to search for connections between concepts like genes, diseases, and drugs, showcasing NLP and information retrieval skills.
Natural Language-to-SQL Copilot
A powerful tool that converts natural language questions into executable SQL queries using AI. Features automatic schema recognition, AI-powered query generation, and supports both OpenAI and HuggingFace models with Spider benchmark integration.
Netflix Tableau Visualization
This interactive visualization provides a comprehensive overview of Netflix's content and viewership trends. Dive into the data to explore how Netflix's vast library captivates audiences worldwide and understand the dynamics driving its success in the global streaming landscape.
Heart Failure Tableau Visualization
This dashboard includes demographic details such as sex and age, along with medical conditions and laboratory measurements, providing a comprehensive view of the characteristics and outcomes within this patient population.
Diabetes Prediction
The project explores the challenges of working with imbalanced data and compares the performance of classical machine learning models against a deep learning approach.
Work Experience
AI Fellow (Contract)
Create prompts and evaluation rubrics used to assess and improve large language model performance.
Machine Learning Engineer Intern
Engineered an end-to-end ML pipeline for an AI tutoring platform, integrating Python-based NLP and text-to-speech services with a Flutter frontend to deliver personalized content with <2s response time in testing.
Data Research Analyst
Led the data lifecycle for interdisciplinary life sciences research, from systematic collection/validation to statistical analysis & Tableau visualization.
General Chemistry Teaching Assistant
Implemented interactive lab sessions for 75+ students, improving comprehension of core chemistry concepts.
Education
M.S. Data Science, Analytics, and Engineering
Concentration in Computing and Decision Analytics.
B.S. Biological Sciences (Biomedical)
Graduated with honors, Dean's List all 8 semesters.
Certifications & Credentials
My commitment to continuous learning and professional development.
Let's Connect
I am actively seeking full-time opportunities and collaborations. Willing to relocate anywhere in the USA. If you have a project in mind or just want to discuss the future of data, I'd love to hear from you.
Or reach out directly
nikitak.nk11@gmail.com