Purvi Verma Profile Picture

My journey as a civil engineering graduate who ventured into artificial intelligence, data science, and computational modeling has shaped me into a researcher eager to bridge disciplines and tackle complex real-world problems.

Computational Insights for Real-World Impact

As a Civil Engineering graduate from IIT Bhubaneswar, my research foundation is defined by its span across Computational Mechanics, Artificial Intelligence, and Data-Driven Modeling. My journey is characterized by interdisciplinary rigor: from developing thermodynamically consistent neural network models for complex materials, to building ML systems for social good, and designing scalable AI safety guardrails in industry. This unique blend of experience in simulation, machine learning, and LLMs drives my pursuit of a PhD, where I aim to integrate diverse methodologies to contribute research that is both scientifically rigorous and societally meaningful.

Explore Research Assets

The Skill Galaxy

The The Knowledge Star (Sun) is orbited by the eight disciplines (Planets) that drive my skillset. Click any planet to see its detailed dossier.

The Knowledge Star

Select a Planet

Click on any orbiting skill to view its specialized knowledge, toolset, and contribution to Scientific Machine Learning.

Skillset:

Professional Timeline

July 2024 - July 2026

Senior Associate Data Science

◦ AI Safety Guardrails: Architected and deployed end-to-end ML pipelines for AI safety, integrating computer vision and LLMs; achieved production-grade reliability with measurable reduction in harmful output rates.

◦ Conversational AI Agent: Designed multi-turn LLM agent with tool calling, modular guardrails, and context-aware workflows; increased task success rate and enabled A/B testing of model variants in production.

◦ Multimodal Content Processing: Built CV + LLM pipeline to extract structured insights from images, videos, and animations; evaluated across 500+ assets for 12 pharmaceutical brands with 90.3% accuracy and 87% semantic consistency.

◦ LLMOps & Monitoring: Championed best practices in model monitoring, versioning, and retraining triggers; presented AI platform strategy to senior stakeholders.

◦ Model Optimization: Explored parameter-efficient techniques (LoRA/PEFT) and vLLM for high-throughput LLM inferencing at scale.

Tap / hover to see more
Doc
May 2023 - May 2024

Research Intern

Topic: Detecting Sexist Texts Targeted at Women in India
Affiliation: SMMME, IIT Bhubaneswar
Department: Institute of Computer Science, University of Tartu
Advisors: Prof. Kisor Kumar Sahu and Prof. Rajesh Sharma
◦ Collected and preprocessed social media data using snscrape & PRAW.
◦ Engineered features: TF-IDF, n-grams, sentiment scores, BERT/RoBERTa embeddings.
◦ Built ML & DL models: Naive Bayes, KNN, ANN, RNN, LSTM, CNN.
◦ Evaluated models using precision, recall, F1-score, ROC-AUC; performed cross-validation & hyperparameter tuning.

Tap / hover to see more
GitHub Doc
March 2023 - May 2023

Artificial Intelligence Intern

◦ Collected and cleaned a large video dataset from YouTube.
◦ Designed a GAN to generate video frames from single images, optimized generator/discriminator for realism.

Tap / hover to see more

Certifications & Courses

Certificate 1

Build Basic Generative Adversarial Networks (GANs)

DeepLearning.AI

Learned the fundamentals of Generative Adversarial Networks, including training basic GANs, understanding their architecture, and practical applications in generating synthetic data.

Certificate 2

Applied Text Mining in Python

University of Michigan

Focused on text mining techniques using Python, including text preprocessing, feature extraction, and sentiment analysis, with hands-on projects using real-world datasets.

Certificate 3

Introduction to Object-Oriented Programming in C++

University of London

Covered core concepts of object-oriented programming, including classes, objects, inheritance, polymorphism, and encapsulation in C++, with practical coding exercises.

Certificate 4

Introduction to Web Development with HTML, CSS, JavaScript

IBM

Gained foundational skills in web development, building responsive websites using HTML, CSS, and JavaScript, and understanding client-side web programming fundamentals.

Certificate 5

Supervised Machine Learning: Regression and Classification

DeepLearning.AI / Stanford Online

Explored supervised machine learning techniques, including regression and classification models, evaluation metrics, and practical implementation using Python and real-world datasets.

Certificate 6

Crash Course on Python

Google

Introduced to Python programming basics, including variables, data types, control structures, functions, and hands-on coding exercises for quick practical understanding.

Personal Projects

Hall of Impact

Agentic AI Hackathon Finalist

Top finalist in the Google Cloud AgenticAI Day Hackathon, demonstrating excellence in rapid LLM workflow and safety development.

General Championship 2024 Winner

Secured 1st place in the Machine Learning-based Real-time Age and Gender Detection competition at IIT Bhubaneswar (2024).

Quantitative & Coding Excellence

Achieved AIR 211 in the JLR Gearshift Code Rally and qualified in the Pre-Regional Mathematical Olympiad (PRMO), 2020.