Welcome to my GitHub profile! I'm passionate about building intelligent systems, particularly in the fields of Artificial Intelligence, Machine Learning, and Generative AI. I specialize in developing Retrieval-Augmented Generation (RAG) systems, Chatbots, and agentic AI workflows that leverage Large Language Models (LLMs) to solve complex, real-world problems. My expertise extends to designing and fine-tuning LLM-based applications for dynamic, context-aware solutions, as well as building multi-agent systems that enable automated reasoning and decision-making. Whether it's creating intelligent chatbots, optimizing RAG architectures, or deploying scalable AI agents, I thrive on pushing the boundaries of what AI can achieve.
- 🔭 I’m currently working as an AI/Machine Learning Engineer at CBase AI, where I build advanced RAG-based chatbots and multi-agent workflows using LangChain and Generative AI.
- 🌱 I’m deeply involved in Generative AI research, focusing on multi-modal LLMs, prompt engineering, and knowledge graph integration.
- 👯 I’m open to collaborating on projects related to AI/ML, NLP, Computer Vision, and Reinforcement Learning.
- 💬 Ask me about RAG systems, LLM fine-tuning, Chatbots, or anything related to Generative AI.
- 📫 How to reach me: balu.koduru99@gmail.com | LinkedIn | Portfolio
- Python, SQL (MySQL, PostgreSQL), MATLAB
- Machine Learning: PyTorch, TensorFlow, Keras, Scikit-learn, Hugging Face, LangChain
- NLP: NLTK, SpaCy, OpenAI GPT, BERT, RoBERTa, LLaMA 2
- Computer Vision: OpenCV, CLIP, LLaVA
- Data Processing: Pandas, NumPy, Spark
- Web Development: FastAPI, ReactJS, Material-UI
- Cloud & DevOps: AWS, GCP, Docker, Kubernetes, Git, GitHub
- Data Visualization: Tableau, Power BI, Streamlit
- Other Tools: ROS, Linux, Anaconda, Jupyter Notebook
- Developed RAG-based Chatbots integrated with SQL Databases, Vector Databases, and Neo4j knowledge graphs for dynamic, context-aware query responses.
- Built multi-agent RAG workflows using LangChain and Generative AI for complex query resolution and automated reasoning.
- Optimized RAG architecture through prompt engineering (Chain-of-Thought, Few-Shot) and Vector Databases, improving retrieval quality by 30%.
- Deployed RAG systems on AWS using Docker and Kubernetes, achieving 99.5% uptime and automated scaling.
- Led the development of EndoAssistant, a Generative AI model for medical surgery analysis, and created the first-ever image-caption dataset in endoscopy.
- Designed a medical knowledge RAG system using LangChain and fine-tuned multi-modal LLMs (LLaVA, CLIP) for evidence-based surgical responses.
- Integrated OpenAI’s Whisper with GenAI-based text correction, reducing transcription errors by 33% in medical speech-to-text pipelines.
- Developed an NLP pipeline using BERT and fine-tuned LLaMA 2 to extract structured data from patient records, reducing documentation time by 60%.
- Implemented Multi-Agent Reinforcement Learning algorithms for autonomous drone fleets, optimizing UAV navigation and coordination.
- Developed CNN-based models for UAV classification and environment analysis, achieving 99.09% and 95.2% accuracy, respectively.
-
Master of Science (MS) in Artificial Intelligence
University at Buffalo, SUNY | May 2024 -
Master of Science (M.Sc.) in Physics & Bachelor of Engineering (B.E.) in Electronics and Instrumentation
Birla Institute of Technology and Sciences Pilani (BITS Pilani) | June 2022
-
Deep Learning Approaches in Modern Thermal Imaging: A Comprehensive Review
IEEE Sensors | Cited 70+ times -
CNN-Based UAV Classification through Time-Frequency Analysis of RC Signal Patterns
IEEE iSES 2022 -
Efficient Deep Learning Architecture for Ambient Sound Detection in Speech Processing
JASA
- Built an AI-powered career platform using ReactJS, Material-UI, and FastAPI, integrating GPT, Anthropic (Claude), and GROQ APIs for automated cover letters, resume optimization, and interview prep.
- Developed a scalable backend on GCP using LangChain, improving candidate-job match rates by 30%.
- Enhanced LLM accuracy by implementing prompt structuring and iterative refinement, reducing hallucinations by 40%.
- Built a RoBERTa-based transformer model for hallucination mitigation, achieving 87.4% accuracy.
⭐️ Feel free to explore my repositories and reach out if you'd like to collaborate or discuss AI/ML projects!