AI/ML Engineer and Data Scientist Analyst

Vandana Pathania

AI/ML/Prompt/NLP Engineer AI/ML Data Analyst

Princeton, NJ 201-***-**** *******.***********@*****.***

linkedin.com/in/vandana-pathania-2bbb7618 github.com/Vanda240 kaggle.com/vandanapathania

SUMMARY

Initiative-taking AI/ML Graduate Researcher with firsthand experience designing, training, and deploying machine learning and deep learning models across NLP, multimodal learning, and structured data domains. Proven ability in transformer-based NLP, including Text-to-SQL generation, topic modeling, and uncertainty-aware regression using PyTorch, TensorFlow, scikit-learn, and LightGBM. Strong background in building reproducible, end-to-end ML pipelines, advanced model evaluation, calibration, and competition-grade modeling with Kaggle-ready submissions.

SKILLS

Languages: Python, SQL, Java, Bash

ML/DL: PyTorch, TensorFlow/Keras, scikit-learn, LightGBM, CNNs, Transformers, Transfer Learning

Computer Vision: CLIP, OpenCV, ResNet, ViT, YOLO

LLMs/GenAI: Hugging Face Transformers, RAG, T5 (Text-to-SQL), LangChain/LangGraph, PydanticAI

MLOps/Backend: Docker, FastAPI, REST APIs, TensorBoard, reproducibility (configs/seeds)

Data/Cloud: Pandas, NumPy, Matplotlib, Plotly, SciPy, Spark, Hadoop, MapReduce, AWS (EC2, S3), Azure

EXPERIENCE

Data Entry Agent (RFP) AI Trainer LinkedIn (Part Time)– May 4 – Jun 4, 2026

Created realistic RFP document intelligence task packages using email threads, PDFs, spreadsheets, and schema-based scenario instructions for AI model training.

Extracted and validated structured RFP fields including deadlines, contacts, submission methods, evaluation criteria, pricing models, compliance requirements, and vendor response sections.

Produced schema compliant JSON ground truth with correct datatypes, picklist IDs, boolean/multiselect values, ISO date formats, and UTC time-zone normalization.

Resolved high-reasoning cases involving conflicting dates, renamed attachments, stakeholder role inference, SOC 2/ISO requirements, and source-grounded field validation.

Kaggle Competitor Kaggle – Online Aug 2025 - Present

Predict the Introverts from the Extroverts: Trained a Random Forest classifier with imputation + binary encoding; achieved 96.7% validation accuracy.

Prediction Interval Competition II (House Prices): Trained LightGBM quantile models (P5/P95) on 200K rows to produce 90% prediction intervals; evaluated using coverage and Winkler score.

Graduate Researcher New Jersey Institute of Technology (NJIT) – Newark, NJ May 2024 – Dec 2025

Developed a rug classification baseline using CLIP ViT-L/14 embeddings + Logistic Regression, achieving 93.5% accuracy on a 200-image test set and improving to 97.6% with TTA on a 1,000-image evaluation.

Developed and benchmarked structured-data prediction models in scikit-learn and TensorFlow/Keras (KNN, Random Forest, Gradient Boosting, LSTM), selecting best models using cross-validation and error analysis; achieved top performance with LSTM (AUC 0.97, Brier 0.06) over classical baselines.

Built a Text-to-SQL explanation layer that parses generated SQL (e.g., SELECT/FROM/WHERE/GROUP BY) and outputs plain-English summaries + rule-based checks, helping non-technical users validate and refine queries.

Implemented and tuned Java Hadoop MapReduce jobs on AWS EC2 (HDFS/YARN), using combiners, split-size tuning, and speculative execution to reduce shuffle/stragglers and achieve ~15% throughput improvement.

Junior Engineer Uttar Gujarat Vij Company Ltd. (UGVCL) – Vijapur, India May 2011 – Jan 2014

Prepared and supported technical documentation for new and changed electrical power connections, ensuring compliance with state utility standards and safety regulations.

Managed and validated government electrical subsidy records, supporting audit readiness and correct disbursement for residential and commercial customers.

Compiled, reviewed, and analyzed monthly and annual inspection reports related to power distribution infrastructure, finding discrepancies and tracking corrective actions.

Coordinated meeting minutes, engineering correspondence, and action-item tracking across field engineers and administrative teams to support prompt execution of distribution activities.

EDUCATION

MS in Artificial Intelligence, New Jersey Institute of Technology (NJIT) – Newark, NJ; GPA: 3.9 / 4.0; Dec 2025

Graduate Certificate in Artificial Intelligence, NJIT – Newark, NJ; GPA: 4.0 / 4.0; Dec 2024

Bachelor of Science in Electrical Engineering, Sankalchand Patel University – India; GPA: 3.0 / 4.0; Jun 2009

VOLUNTEER EXPERIENCE

AI/LLM Engineer (Project) Think Round / Zoe Agent – Remote Jan 15 2026 – Apr 15 2026

Analyzed Zoe, a trauma-informed conversational AI agent using plugin-based agents, centralized

CortexFlow orchestration, LangGraph workflows, and OpenAI/Anthropic provider adapters.

Built confidence-scoring logic combining rule-based checks with LLM evaluation for relevance, helpfulness,

accuracy, and completeness; used threshold-based retries and fallback responses for safer outputs.

Implemented RAG-style knowledge indexing over project context files and debugged path/configuration issues in

backend workflow and knowledge-index builder modules to improve grounded response reliability.

Added guardrails, response validation, and debugging recommendations to reduce low-confidence, verbose, or

unsafe responses in a production-style FastAPI backend environment.

CERTIFICATIONS & TRAININGS

• Certifications: Gen AI Foundations & Prompt Engineering, Miro Fundamentals, Alteryx

• Trainings: Active Listening, Adaptability Mindset, Creative Thinking, Critical Thinking, Emotional Intelligence, Making Quick Decisions

Contact this candidate