Hi, I'm Naveen Addanki

AI/ML Engineer

Building intelligent AI systems with expertise in LLMs, RAG, and Machine Learning

About Me

AI/ML Engineer with expertise in building intelligent systems using LLMs, deep learning, and advanced machine learning techniques. Currently pursuing Master of Science in Data Science at Indiana University Bloomington with a 3.9/4.0 GPA.

Specialized in developing agentic RAG systems, fine-tuning large language models, and deploying production-ready AI applications. Experienced in building real-time data processing pipelines, implementing MLOps best practices, and creating scalable ML solutions across AWS, GCP, and Azure platforms. Passionate about leveraging AI to solve complex problems and drive innovation.

AI & LLM Engineering

Expert in fine-tuning, post-training, RAG systems, and prompt engineering with LangChain & Hugging Face

Machine Learning

Proficient in PyTorch, TensorFlow, Scikit-learn, and building end-to-end ML pipelines

Cloud & MLOps

Experience with AWS (SageMaker, Bedrock, Lambda), GCP (Vertex AI), and Azure for ML deployment

Data Engineering

Skilled in building scalable data pipelines, ETL workflows, and real-time processing systems

Experience

My professional journey and key accomplishments

AI/ML Engineering Resident

Headstarter

Sep 2025 - Present
Remote
  • Built 4+ machine learning, AI-engineering and full-stack projects in fast-paced software team environments
  • Developed 5+ neural networks in Python, 11 apps in TypeScript on AWS/Vercel with dev and production environments
  • Implemented LLM-chaining, hyperparameter tuning, fine-tuning on 10+ LLM models controlling for latency & accuracy

Graduate AI/ML Research Engineer

Indiana University

Jan 2025 - Aug 2025
Bloomington, IN
  • Enabled analysts to receive real-time, data-driven insights by developing an agentic RAG system and embedded 5,000+ SEC filings and financial news articles for fast, hybrid retrieval through a real-time data ingestion layer, vector database, integrated FastAPI web app
  • Improved chatbot answer accuracy and boosted analyst productivity by reducing manual research time by 10x and interactive dashboard
  • Fine-tuned SD XL1.5 and LLaMA-3-8B using LoRA for origami image generation and review classification improving model accuracy to 85% through DPO, RLHF methods on multiple GPUs and quantized using the GPTQ for inference
  • Deployed end-to-end applications for AI image synthesis, enabling user interaction and real-time generation via Python backend service
  • Achieved 95% labeling accuracy and supported large-scale ML data needs by managing 50+ annotators and curating 250K+ labels

Graduate Generative AI Engineer

Indiana University, Eskenazi School of Art, Architecture + Design

May 2024 - Dec 2024
Bloomington, IN
  • Built real-time Voice AI using RAG, deployed hybrid semantic search pipelines, improving precision, reducing irrelevant responses by 30%
  • Engineered graph-based reasoning layer with Neo4j for structured data augmentation and memory-managed dialogue systems supporting 3000+ conversations per day with long- and short-term context retention
  • Integrated vector database to embed 45+ hours of transcripts, cutting retrieval latency by 40% and boosting context relevance for voice
  • Dockerized, orchestrated CI/CD pipelines through cross-functional collaboration with DevOps teams and domain experts, for seamless deployment

Machine Learning Engineer

Amgen (Client)

Sep 2021 - Jul 2023
Remote
  • Developed customer segmentation models improved business KPIs by 6%, feature-engineered and deployed in cloud using REST APIs
  • Engineered CI/CD pipelines to automate model training/testing, ensuring reproducibility and integration with DevOps teams for production
  • Built scalable Informatica ETL workflows to transform 7M+ unstructured records daily into analytics-ready formats for modeling & reports
  • Supported model and data drift monitoring to identify performance issues and assist with retraining, ensuring model reliability
  • Created Power BI dashboards and Power Automate workflows used by 50+ stakeholders, accelerating decision-making processes

Skills

My expertise and technical proficiencies

AI/LLM

Fine-tuning
Post Training
SFT
Crew AI
Prompt Engineering
Agents
CUDA
vLLM
LangChain
Hugging Face

Machine Learning

Python
PyTorch
TensorFlow
Keras
Scikit-learn
Transformers
NLTK
DVC
MLflow

Data & Databases

MySQL
MongoDB
NoSQL
Qdrant
FAISS
Pinecone
Neo4j
Chromadb

Cloud Platforms

AWS (S3, EC2, Lambda, SageMaker, Bedrock)
Azure
GCP (BigQuery, Vertex AI)
Snowflake

MLOps & DevOps

Docker
Kubernetes
Jenkins
CI/CD
Airflow
Git
GitHub
JIRA

Data Visualization

PowerBI
Tableau
Matplotlib
Data Analytics

Projects

A selection of my AI/ML and data science projects

Real-Time Data Processing & Anomaly Detection

Built a real-time data processing pipeline with Apache Flink, analyzing streaming order and payment data, improving event processing latency by 40% and reducing fraudulent event detection time from 5 mins to 10s.

Apache FlinkReal-time ProcessingAnomaly Detection

Hotel Reservation Cancellation Prediction

End-to-end cancellation prediction system using LightGBM achieving 92% accuracy, with MLflow for experiment tracking & model versioning. Deployed through Jenkins CI/CD with web app interface.

LightGBMMLflowJenkins

Secure LLM Fine-tuning and Deployment

Leveraged Unsloth and QLoRA to fine-tune LLaMA-3.2, reducing VRAM usage by 70% and accelerating training by 2x. Strengthened model robustness through adversarial training via TextAttack, lowering attack success rates by 11%.

LLaMAQLoRAUnsloth

Hybrid Anime Recommendation System

Built a recommendation engine combining collaborative filtering and content analysis, processing 100K+ entries through GCP. Implemented MLOps with COMET-ML, DVC, and Jenkins CI/CD, deploying on Google Kubernetes Engine.

GCPCollaborative FilteringMLOps

Brain Tumor Classification

Used neural networks in Python to classify 1000 MRI scans into 3 types of possible brain diseases with a custom model. Generated multimodal MRI reports for neurosurgeons in under 200ms after image classification.

Deep LearningComputer VisionMedical AI

Agentic RAG System for Financial Analysis

Developed an agentic RAG system embedding 5,000+ SEC filings and financial news articles with hybrid retrieval through vector database and FastAPI web app, reducing manual research time by 10x.

RAGLangChainVector DB

Education

My academic background and qualifications

Master of Science - Data Science

Indiana University Bloomington

GPA: 3.9/4.0 | Aug 2023 - May 2025

Pursuing advanced studies in Data Science with focus on AI/ML, deep learning, and statistical modeling. Coursework includes Reinforcement Learning, Statistics, Applied Algorithms, Computer Vision, Data Analytics, Data Structures, and Deep Learning.

Key Coursework:

Reinforcement LearningDeep LearningComputer VisionApplied AlgorithmsStatisticsData Analytics

Blog

Thoughts, tutorials, and insights on software development

Visit My Blog

I write about AI/ML, LLMs, Deep Learning, Data Science, and other cutting-edge technologies on my Hashnode blog. Check out my latest articles and tutorials.

Visit Blog

Get In Touch

Have a project in mind or want to discuss opportunities? I'd love to hear from you!