Skip to content
AboutSkillsExperienceProjectsEducationContact
Download resume

Jersey City, NJ

Vraj Shaileshbhai Patel

Software & Machine Learning Engineer

Building reliable software and data-driven solutions, end to end.

View ProjectsGet in Touch
Photo of Vraj Shaileshbhai Patel

About

I'm a Computer Science M.S. candidate (graduating May 2026) who enjoys building and shipping software across the full lifecycle — from working with data and designing solutions to deploying them in the cloud. I like turning messy problems into clean, reliable systems.

I work primarily in Python and SQL and care about reproducible, version-controlled workflows and writing code that holds up. I'm comfortable picking up new tools quickly and translating technical work for both technical and non-technical audiences.

Driven by curiosity and a love for building. Always looking to learn something new and take on the next challenge.

Skills

Machine Learning

scikit-learnXGBoostRandom ForestLogistic RegressionCross-ValidationHyperparameter TuningROC-AUC / Model EvaluationFeature Selection

Data Analysis

Python (Pandas, NumPy)EDAStatistical AnalysisMatplotlib

Healthcare Data

Clinical Dataset AnalysisConfidential Health-Data HandlingMedical-Imaging Research

Programming Languages

JavaScriptCC++JavaPython

Web & Frameworks

ReactHTMLCSSTailwind CSSFastAPINode.js

SQL & Databases

SQLPL/SQLMySQLPostgreSQLMongoDB

Tools

GitGitHubDockerVS CodeJupyterLinux CLIExcel

Spoken Languages

EnglishHindiGujarati

Experience

Sep 2025 - Present

Graduate Researcher

New Jersey Institute of Technology · Newark, NJ

Clinical Data Analysis, PLCO Cancer-Screening Trial. Advisor: Prof. Arashdeep Kaur

  • Analyze the PLCO cancer-screening dataset in Python (Pandas), running EDA and statistical testing to rank candidate predictors for downstream risk modeling.
  • Process confidential patient health and medical-imaging data under strict privacy and governance standards, keeping records secure and analysis-ready.
  • Build preprocessing and feature-engineering pipelines, cutting redundant variables by 30% through data-driven feature selection.
  • Evaluate models with AUC, F1, and Brier score, and share findings through a standardized, reproducible workflow for both technical and non-technical stakeholders.

Jan 2023 - Apr 2023

Web Systems Design & Development Intern

Mind Inventory · Ahmedabad, India

  • Built and maintained scalable, database-driven web systems using PHP, MySQL (XAMPP), and front-end frameworks.
  • Worked in an agile team on relational databases and responsive UI across 3+ platforms, improving page-load speed by roughly 30%.
  • Contributed to QA testing and bug-fix cycles, strengthening code reliability.

Jun 2022 - Jul 2022

Web Developer Intern

EliteEvince Technologies · Ahmedabad, India

PHP & MySQL

  • Translated client requirements into deliverables under tight timelines.
  • Resolved UI and documentation issues across 2 internal projects, improving maintainability and usability.

Projects

GraphRAG — Multi-Cancer Prediction Research
PythonTypeScriptReactFastAPINeo4jLangChainClaudeGeminiVercelRender
  • Knowledge-graph question-answering system over sixteen peer-reviewed cancer-prediction research papers: built a GraphRAG pipeline that combines vector search with Neo4j graph traversal, so answers are grounded in explicit graph facts rather than opaque text similarity.
  • Designed the ingestion pipeline with LangChain's LLMGraphTransformer (Claude Haiku) to extract entities and relationships from paper chunks into a Neo4j knowledge graph, linking each entity back to its source chunk for traceability.
  • Combined Gemini embeddings with a Neo4j vector index for chunk retrieval, then traversed one hop out from retrieved chunks' entities to pull in connected relationships, letting Claude answer multi-hop questions plain RAG can't (e.g. connecting a dataset to a model never mentioned in the same sentence).
  • Shipped a full-stack app end to end: FastAPI backend, React/TypeScript/Tailwind frontend with explainable Entity → RELATIONSHIP → Entity fact display, deployed across Vercel, Render, and Neo4j Aura with environment-driven CORS and build-time API wiring.
  • Solved a real production constraint by switching from a local HuggingFace embedding model to the Gemini embeddings API after memory limits crashed the deploy target, building a `--reembed` mode to re-embed the full corpus at a new vector dimension without re-running costly LLM extraction.
GitHubLive Demo
Lung Cancer Risk Prediction
Pythonscikit-learnXGBoostLightGBMCatBoostpandasMatplotlib
  • Data efficiency and stability analysis: designed a super-stacking ensemble of six base learners (Logistic Regression, Random Forest, Gradient Boosting, XGBoost, LightGBM, CatBoost) with a logistic-regression meta-learner to predict lung cancer risk from 15 clinically interpretable variables in the PLCO trial.
  • Ran a data-efficiency study across training fractions from 100% down to 10%, evaluated against a fixed, stratified 20% test set with 5 random seeds per fraction, reporting mean and standard deviation for reproducibility.
  • Held AUC-ROC near 0.836 with only a ~0.005 drop at 10% training data, keeping recall around 0.88.
  • Mitigated severe class imbalance via class_weight / scale_pos_weight and tuned the decision threshold on validation only, keeping the operating point stable near 0.30 without test-set leakage.
GitHub
VolGuard
Pythonscikit-learnXGBoostSQL (SQLite)AWS LambdaAWS S3TerraformTableau
  • End-to-end stock volatility forecasting: built an ML pipeline forecasting 5-day realized volatility for five large-cap stocks, from raw prices through SQL feature engineering, a validated model, AWS Lambda, and a live Tableau dashboard.
  • Engineered features entirely in SQL (daily returns, 5/10/30-day rolling volatility, forward-5-day target) with strict chronological splits to prevent data leakage.
  • Validated with walk-forward testing across 4 out-of-sample folds, benchmarking Linear Regression, Random Forest, XGBoost, and GARCH(1,1) against a naive baseline, and shipped the simplest model, cutting RMSE by roughly 20% versus baseline.
  • Deployed to AWS with Terraform (S3 + Lambda, API-key auth), serving predictions via CLI/SDK that feed the public dashboard.
GitHubLive Dashboard
Mini Amazon
ReactTailwind CSSFastAPIMongoDBJWTWebSockets
  • Full-stack e-commerce app: built a scalable platform for 100+ users with product search, cart, and order tracking.
  • Developed an admin dashboard supervising 500+ products with real-time order updates.

Education

New Jersey Institute of Technology

M.S. in Computer Science

GPA: 3.80 / 4.0

Sep 2024 - May 2026

Newark, NJ

Gujarat Technological University (SAL Institute)

B.E. in Computer Engineering

GPA: 8.13 / 10

Sep 2019 - May 2023

Ahmedabad, India

Open to opportunities where I can build meaningful things and learn along the way.

Reach out through any of the channels below. I usually reply within a day or two.

pvraj0019@gmail.com+1 (605) 212-0452LinkedInGitHub