AI/ML/DL & Backend Engineering Professional | MSc in Artificial Intelligence (Distinction) from University of London

"Kush" works just as well if we're being casual 😊

🔹 Currently focused on refreshing and leveling up my Backend Engineering skills, especially in Microservice Architecture and Containerization using Docker.
🔹 Parallelly continuing my journey in AI/ML research and experimentation with a strong focus on practical applications.
🔹 3+ years in Java, Spring Boot, Microservices, and Cloud Computing
🔹 Researching Uncertainty in LLM Responses using Conformal Prediction & Active Learning

University of London, London, UK
📅 Sept 2023 – Sept 2024 | 🎓 Distinction (Overall Aggregate: 80.4%)

Thesis: "Exploration of Transfer Learning and Hyperparameter Optimization in Deep Learning"

• Implemented and compared multiple pretrained architectures (VGG, ResNet, etc.) on image datasets
• Conducted experiments on activation functions, learning rates, batch sizes, and regularization
• Explored active learning strategies for data-efficient training and transfer learning for low-data tasks
• Evaluated models for convergence speed, generalization, and impact of fine-tuning depth

Visvesvaraya Technological University, Belagavi, India
📅 Aug 2017 – Sept 2021 | 🎓 CGPA: 7.95/10 (Distinction)

Project: "Automated Detection and Recognition of License Plates (ANPR System)"

• Developed a YOLOv3-based object detection pipeline for license plate localization
• Integrated OpenCV and Tesseract OCR to extract and recognize alphanumeric characters
• Achieved high recognition accuracy on vehicle images under varying lighting/angles
• Published in IJSREM: 🔗 View Paper

Backend Engineering: Java (Core & Advanced), Spring Boot, Hibernate, REST APIs, Microservices
AI/ML & Research: Deep Learning, NLP, LLMs, Conformal Prediction, Cross-Conformal Prediction, Active Learning, TensorFlow, PyTorch
Databases: MySQL, DynamoDB, PostgreSQL
Cloud & DevOps: AWS (EC2, ECS, DynamoDB), Docker, Kubernetes, CI/CD (Jenkins, GitLab CI)

Title: "Uncertainty Estimation in LLM Responses using Conformal Prediction & Active Learning"
🔹 Goal: Improve reliability of LLM-generated responses in high-stakes applications
🔹 Approach: Apply Conformal Prediction & Cross-Conformal Prediction techniques to assess uncertainty in AI-generated text
🔹 Future Application: Integrating this method into AI-assisted PR Code Reviews & Developer Scoring
🔗 View Repository

🔥 AI-Powered PR Code Reviewer (Open for Collaboration)
🔹 Concept: Using LLM + Conformal Prediction + Active Learning to determine uncertainty in AI-generated code reviews
🔹 Goal: Develop an intelligent AI assistant for scoring PR quality & developer contributions
🔹 Status: Still in idea phase—actively exploring core concepts before starting development
🔹 Looking for Collaborators! If you're interested in AI + Software Engineering, let’s connect
🔹 Tech Stack (tentative): Java, Spring Boot, OpenAI API, PyTorch, React.js

⭐ Fun Fact: Always experimenting with new tech — whether it's LLMs, microservices, or deep learning models! 🚀

• LinkedIn: b-s-kushal
• Email: kushal.basapathi@gmail.com