Pavan Sai Reddy Pendry

Skills & Tools

Languages

Python, C++, JavaScript, SQL, Java, C

Frameworks & Libraries

React.js, FastAPI, Flask, Pandas, TensorFlow, PyTorch, Apache Kafka, OpenAI SDK, LangChain, Hugging Face Transformers, BeautifulSoup, RapidFuzz

Databases

PostgreSQL, MongoDB, Redis, Supabase, Vector Databases (ChromaDB, pgvector)

Cloud & DevOps

AWS (S3, Lambda, ECS, Step Functions, API Gateway), Docker, Git, GitHub Actions, Vercel, Render

AI/ML Concepts

RAG, RLHF, LLMs, NLP, Vector Embeddings, Semantic Search, Query Routing, ETL Pipelines, Prompt Engineering

Tools

Power BI, Postman, Jupyter Notebook

Experience

Software Engineer (GenAI) – University of Kansas

Oct 2024 – Present

Live Demo GitHub

Building BabyJay, an AI-powered campus assistant serving 9,500+ knowledge documents with 82.4% user approval rate.

Developed fullstack application using React.js frontend with dark mode UI, Google OAuth, and mobile-responsive design, paired with FastAPI backend handling JWT authentication and streaming responses.
Engineered RAG-based retrieval system with 4 specialized retrievers (faculty, course, campus, vector) achieving 35x speedup (700ms to 20ms) through hybrid search combining vector embeddings with database indexing.
Built intelligent query routing pipeline with intent classifier, query preprocessor for synonym expansion and typo correction using RapidFuzz, and embedding-based intent detection with regex fallback.
Implemented real-time enrollment data integration by building web scrapers for classes.ku.edu, enabling live seat availability queries across 7,355 courses with 2-layer semantic caching.
Created ETL pipelines using Python and shell scripts to process 26,453 raw records into PostgreSQL across 72 department categories with automated data ingestion workflows.
Designed feedback analytics system with thumbs up/down collection and built RLHF optimization pipeline that learns from user feedback to dynamically enhance system prompts.
Deployed containerized backend using Docker with CI/CD automation using GitHub Actions. Frontend hosted on Vercel, backend on Render with custom domain configuration.

React.js FastAPI PostgreSQL Docker RAG RLHF OpenAI Supabase Python GitHub Actions

Graduate Teaching Assistant – University of Kansas

Jan 2025 – Dec 2025

Delivered interactive lectures and hands-on workshops to 20 students on data analytics covering collection, preprocessing and visualization using Python libraries such as NumPy, Pandas, Matplotlib and Seaborn.
Taught supply chain analytics topics including customer segmentation, warehouse optimization and demand forecasting.
Mentored students 1:1, contributing to a 20% increase in average assessment scores.
Leading recitations, grading and office hours for ~50 students in MATH 101 (Aug 2025 – Present).

Python Pandas Power BI Data Analytics

Software Engineer – Amrita Vishwa Vidyapeetham

Jan 2023 – May 2024

Engineered DishKit, a fullstack web application using React.js frontend and Flask backend, integrating LSTM-based next-word prediction and nutrient analysis models. Built interactive dashboards with PowerBI helping nutritionists create 50+ personalized meal plans weekly.
Developed real-time predictive UI using Bidirectional LSTM (491K parameters) with Python and TensorFlow, boosting dish selection efficiency and improving user engagement by 3x.
Designed and deployed SweepSpot, a fullstack waste management platform using React.js frontend and FastAPI backend, serving 500+ users across 10 municipalities with geolocation tracking, image uploads, and reporting dashboards built with PowerBI.
Built scalable data pipelines using Python and PostgreSQL handling 1,000+ geo-tagged reports. Implemented Haversine-based distance calculations for route optimization achieving 90% accuracy.

React.js Flask FastAPI PostgreSQL TensorFlow Power BI MongoDB

Projects

QuickDrop

Feb 2025

Built a file transfer tool for Mac and Android after getting frustrated with existing solutions. No cloud uploads, no accounts, no ads.

Developed a Flask-based web server with QR code connectivity, chunked uploads for large files (3GB+), and a drag-and-drop browser UI. Added ADB integration for USB transfers hitting 30+ MB/s.

Open source project that gained traction on LinkedIn. Documented the entire build process including bugs and fixes.

Python Flask JavaScript ADB HTML/CSS

GitHub

Market Data Service

May 2025 – Jun 2025

Implemented a real-time stock quote pipeline using the Finnhub API and PostgreSQL via SQLAlchemy, caching hot prices with Redis. Achieved 99.99% data availability and reduced database query response times by 60%.

Designed and deployed a Kafka streaming pipeline to publish price events and compute five-point moving averages in real time.

Engineered a FastAPI microservice with REST endpoints, containerized the system with Docker Compose and automated CI/CD workflows.

FastAPI Kafka PostgreSQL Redis Docker SQLAlchemy

GitHub

AI-Generated Text Detection System

Feb 2025 – Apr 2025

Collected and annotated a 100,000-sample dataset (50K human, 50K AI) for text classification, achieving a 78% baseline accuracy. Fine-tuned a LLaMA model on AWS EC2, improving detection accuracy to 88%.

Deployed a serverless detection API on AWS Lambda with S3 storage and EC2 inference, supporting 2,000 concurrent requests at under 200ms latency.

Integrated the model into a React/Node.js web app with five input sources, reducing manual review time by 40%.

LLaMA AWS Lambda AWS EC2 React Node.js S3

Emotion Aware Chatbot

Nov 2024 – Dec 2024

Tailored a hybrid BERT-GPT framework for real-time emotion recognition, surpassing 89% accuracy across six emotion classes.

Optimized the inference microservice to handle 500 concurrent requests at under 150ms latency.

Created a reactive Streamlit UI hosted on Heroku for 100 daily beta users, reducing average response latency by 25% and cutting load times to 300ms.

BERT GPT Streamlit Heroku Python

Hello, I'm Pavan Sai Reddy Pendry

About Me

Education