Hi, my name is
Naga Vinay Avvaru.
I build intelligent systems.
I'm an AI/ML & Agentic AI specialist focused on building computer vision systems, autonomous agent workflows, and production-grade MLOps pipelines. Currently pursuing B.E. in ECE at St. Joseph's Institute of Technology.
About Me
Hi! I'm Vinay, a B.E. student in Electronics and Communication Engineering at St. Joseph's Institute of Technology (2023–2027), with a deep passion for AI/ML and intelligent systems. My journey started when I first trained an image classifier — and I've been hooked on machine learning ever since.
I specialize in building computer vision systems and Agentic AI workflows. I've interned at Spheruler Solutions, where I improved military-grade camera calibration accuracy by 3x, and at Zeka pvt Solutions, where I built AI agent pipelines with robust prompt injection defenses (47% security improvement).
When I'm not coding, I compete in hackathons — most recently winning National 2nd place for an AI-powered elephant detection system for wildlife conservation.
Here are a few technologies I've been working with recently:
- Python
- TensorFlow / PyTorch
- LangChain / LangGraph
- OpenCV / YOLOv8
- FastAPI
- GCP / Docker
Where I’ve Worked
Vision Systems Intern @ Spheruler Solutions
March 2025 - May 2025
- Engineered a military-grade camera calibration system for boresight alignment, identifying intrinsic and extrinsic parameters using geometric calibration techniques
- Reduced reprojection error from 0.15px to ~0.05px, improving alignment accuracy by 3x for high-precision optical targeting applications
- Developed Python-based calibration tooling with OpenCV to automate the camera parameter extraction pipeline, reducing manual calibration time significantly
- Collaborated with senior engineers to validate calibration outputs against physical measurement benchmarks in real-world scenarios
Some Things I’ve Built
Featured Project
CODEPRO — AI Plagiarism Detection
A modular, end-to-end code plagiarism detection system that goes beyond simple text matching. Implements the Winnowing algorithm with k-gram fingerprinting and Rabin-Karp rolling hashes to accurately detect copied code. Features a robust normalization pipeline to catch variable renaming, whitespace changes, and comment removal — the tricks students typically use to evade naive detectors.
- Python
- Winnowing Algorithm
- Rabin-Karp Hashing
- FastAPI
- Streamlit
![CODEPRO — AI Plagiarism Detection]()
Featured Project
Regression-ML-EndtoEnd
A production-grade, end-to-end MLOps pipeline designed for real-world regression tasks. Features a modular architecture for automated feature engineering, XGBoost model training, and hyperparameter tuning via Optuna. The trained model is containerized and deployed as a FastAPI service on Google Cloud Run, with Supabase used for persistent prediction storage and experiment tracking.
- Python
- XGBoost
- Optuna
- FastAPI
- GCP Cloud Run
- Supabase
![Regression-ML-EndtoEnd]()
Featured Project
Fish Detection — YOLOv8
A custom-trained YOLOv8 real-time object detection model for fish species identification and tracking in underwater imagery. Optimized to handle challenging conditions including low-light environments, high turbidity, and partial occlusion. The open-source project gained 50+ GitHub stars and has been referenced by other researchers working on marine biology computer vision tasks.
- Python
- YOLOv8
- PyTorch
- OpenCV
- Ultralytics
![Fish Detection — YOLOv8]()
Other Noteworthy Projects
view the archiveRAG Teacher Voice Agent
A RAG-powered multilingual voice assistant built for educational navigation and student support. Combines retrieval-augmented generation with a voice interface to answer queries from a knowledge base. Supports 10+ languages, built as part of a top-16 SIH internal hackathon finishing project (PreserVion).
REST API Microservice
A production-ready REST API microservice built with FastAPI, following clean architecture and microservice design patterns. Features containerized deployment with Docker, structured logging, error handling middleware, and documented endpoints — a reusable foundation for backend services.
AI Trip Planner
An AI-powered travel planning assistant that generates personalized trip itineraries using large language models. Features include destination research, budget estimation, day-by-day schedule generation, and multi-city routing — all through a conversational Streamlit interface.
Certifications
What’s Next?
Get In Touch
Although I’m not currently looking for any new opportunities, my inbox is always open. Whether you have a question or just want to say hi, I’ll try my best to get back to you!
Say Hello