Pragnya Ambekar

Projects

Master's Thesis TypeScript Static Analysis

React Architectural Anti-Pattern Detector

A TypeScript CLI tool that detects six React architectural anti-patterns by traversing the AST using the TypeScript Compiler API. Built as my Master's thesis at UW Bothell.

Detects God Components using a combined metric of LOC, JSX count, and hook count, achieving 98.04% precision
Formally evaluated by 3 independent reviewers across 71 stratified components, achieving 86.39% average precision
Modular architecture makes it easy to extend with new detection rules

Accessibility Web App AngularJS

SwarajyAbility – Job Portal for People with Disabilities

A comprehensive job portal designed specifically for people with disabilities. This platform matches candidates with suitable job opportunities.

Accessible UI with screen reader compatibility
Smart job matching algorithm
Resources for workplace accommodations

NLP Toxicity Detection HuggingFace

Detoxify Comment Analyzer

A toxicity detection tool for analyzing online comments using the Detoxify NLP model from Hugging Face Transformers.

Uses pretrained Detoxify model for real-time analysis
Detects toxic, severe toxic, obscene, threat, insult, and identity hate categories
Clean UI for text input and categorized predictions

Machine Learning Data Analysis Python

Visual Defect Detection in Manufacturing

Built an AI-powered system to detect manufacturing defects using transfer learning with ResNet-50 and Grad-CAM for visual explanations.

Achieved 96.67% accuracy on single-category defect detection
Used Grad-CAM to highlight defect areas visually
Deployed as an interactive Streamlit web app

NLP Email Security Python

Inbox Guard: Spam & Phishing Detection

Developed a lightweight email security tool to classify and block spam and phishing emails using machine learning.

Used NLP and text classification with scikit-learn
Trained on labeled spam/phishing datasets
Deployed as a Flask web app for testing and demos

Machine Learning Network Security React Flask

CyberML — Network Intrusion Detection

A full-stack app that detects network intrusions using machine learning. Five algorithms are trained on the KDD Cup 99 dataset and served through a Flask REST API, with a React frontend for real-time prediction and model comparison.

Classifies network traffic into 5 categories: NORMAL, DOS, PROBE, R2L, U2R using Decision Tree, Naive Bayes, MLP, and K-Means
Interactive prediction form with preset attack scenarios and algorithm accuracy comparison charts
Client-side auth with SHA-256 password hashing and a clean React dashboard for analysis

Experience

Jun 2025 – Oct 2025

Software Engineering Intern

Secure AIs • Seattle, WA

Built the document upload and sanitization flow using React and FastAPI, making sure LLM outputs were validated for safety before anything reached the user.
Built an AI chat interface with prompt injection guards and audit logging so every interaction was secure and traceable.
Worked with backend engineers to define API contracts early in the process, which kept integration smooth across upload, sanitization, and review stages.
Made sure sensitive document data never ended up in browser storage or client logs by implementing secure state cleanup and safe file handling across the frontend.

React FastAPI LLM AI Security REST APIs

Oct 2024 – May 2025

Graduate Research Assistant

University of Washington – Bothell

Looked into why toxicity models were misclassifying messages across 100K+ Twitch chats. Found it came down to messy data and class imbalance. After fixing both and tuning BERT, RoBERTa, and DistilBERT, F1 went from 0.74 to 0.89.
The preprocessing pipeline was taking too long so I rewrote it using batch processing, multiprocessing, and vectorization. Cut the runtime by 40% which made running experiments a lot faster.
Set up a consistent ML workflow with PyTorch and Hugging Face so it was easy to swap models and compare results across experiments.

NLP Python BERT RoBERTa PyTorch Hugging Face scikit-learn

Jan 2023 – Sep 2024

Software Development Engineer

Byteridge • Hyderabad, India

Contract search was slow, so I used the MongoDB Explain profiler to find the problem and redesigned the indexes around real usage patterns. Search latency dropped by 25 to 40 percent.
Worked across the full stack on CLM features including dashboards, user management, and audit logs using Java, Spring Boot, Node.js, Angular, and TypeScript.
Built a Signature Adapter microservice using the Factory pattern so the platform could work with DocuSign or Adobe Sign interchangeably and add new providers later without changing core code.
Found and fixed 50+ WCAG issues across the UI covering keyboard navigation, screen readers, ARIA, focus management, and color contrast. The platform passed its accessibility recertification.

Java Spring Boot Angular TypeScript Node.js MongoDB WCAG

Jan 2023 – Jun 2023

Software Development Engineer Intern

Byteridge • Hyderabad, India

Worked on SwarajyAbility, a job portal built for people with disabilities. Spent a lot of time on accessibility, making sure everything worked with screen readers, keyboard navigation, and met WCAG standards.
Built the job matching flow and accommodation resource pages with a focus on how the experience actually feels for someone using assistive technology.
Also helped with a Spring Boot microservices migration on the backend to improve the overall architecture.
Cleaned up Hibernate entity mappings and tuned PostgreSQL queries which reduced query failures by around 35 percent.

AngularJS Java Spring Boot PostgreSQL Hibernate WCAG

Jun 2022 – Nov 2022

Full Stack Developer Intern

NFC Solutions • Hyderabad, India

Supported backend development using PHP and Zend Framework for web-based applications
Reduced backend errors during testing by 20-25% through improved request-handling logic
Decreased data inconsistencies by 30% through SQL database optimization using PhpMyAdmin
Learned React fundamentals and full-stack application architecture

PHP Zend Framework SQL React PhpMyAdmin

Aug 2021 – Nov 2021

Full Stack Developer Intern

GeoGo Solutions • Remote

Supported development and maintenance of 8-10 web pages for internal tools and external platforms
Reduced mobile-specific layout issues by 30-40% through responsive design improvements
Fixed 15-20 frontend issues including misaligned elements, broken links, and form behavior bugs
Gained exposure to full SDLC including requirements, implementation, testing, and feedback loops

HTML CSS JavaScript Responsive Design

Apr 2021 – Aug 2021

Head of Web Development

Edam IARE (College Tech Club) • Hyderabad, India

Led four-member web development team for college tech club's official website
Reduced update turnaround time by 50% from 1-2 days to a few hours
Mentored 12 junior students in web development, with 9 securing internships
Designed and developed responsive landing page using HTML5 and CSS
Deployed website using Netlify, gaining hands-on deployment experience

Leadership HTML5 CSS Mentoring Netlify

Achievements

Best Performer Award (2×) – Byteridge

Awarded Best Performer for two consecutive quarters in recognition of consistent performance, ownership of features, and high-quality contributions as a Software Development Engineer.

Runner-Up - National Web Development Contest

Recognized at a national level for the Budget Tracker full-stack project developed during the Internshala program, competing against developers nationwide.

Topper of the Department - Information Technology (IARE)

Awarded for academic excellence and top performance in the Information Technology department throughout the B.Tech program.

1st Place - Lexicon Hackathon (IARE)

Secured first position in a college-level hackathon for building a technical solution under time constraints, demonstrating rapid prototyping and problem-solving skills.

My Skill Set

Languages

Frontend

Backend

Databases

Machine Learning

Cloud and DevOps

Projects

React Architectural Anti-Pattern Detector

SwarajyAbility – Job Portal for People with Disabilities

Detoxify Comment Analyzer

Visual Defect Detection in Manufacturing

Inbox Guard: Spam & Phishing Detection

CyberML — Network Intrusion Detection

Experience

Software Engineering Intern

Secure AIs • Seattle, WA

Graduate Research Assistant

University of Washington – Bothell

Software Development Engineer

Byteridge • Hyderabad, India

Software Development Engineer Intern

Byteridge • Hyderabad, India

Full Stack Developer Intern

NFC Solutions • Hyderabad, India

Full Stack Developer Intern

GeoGo Solutions • Remote

Head of Web Development

Edam IARE (College Tech Club) • Hyderabad, India

Achievements

Best Performer Award (2×) – Byteridge

Runner-Up - National Web Development Contest

Topper of the Department - Information Technology (IARE)

1st Place - Lexicon Hackathon (IARE)

Education

Master of Science (M.S.)

Bachelor of Technology (B.Tech)

Let's Chat

Pragnya Ambekar