Niranj Patel
Niranj Patel
About
Driven AI Engineer passionate about building and deploying advanced Large Language Model (LLM) and Generative AI (GenAI) solutions that solve real business problems. Specialized in designing scalable, end-to-end machine learning systems, transforming ideas from prototype to production with measurable results.
LLM Ops: Prompt engineering, Retrieval-Augmented Generation (RAG), pipeline orchestration, custom LLM fine-tuning, deploying models at scale (GPT-4, Claude, Llama)
MLOps & CI/CD: MLflow, Docker, AWS SageMaker, Lambda, Kubernetes
GenAI apps: Chatbots, Document search(Docubot), E-commerce Gen AI, medical cost prediction
Tech stack: Python, PyTorch, TensorFlow, LangChain, RAG, Gen AI, ChromaDB, FAISS, OpenAI APIs, FastAPI, AWS (Bedrock, SageMaker)
- Built multi-agent RAG pipelines powering enterprise-grade chatbots
- Delivered Docubot, an intelligent document Q&A assistant using multi-agent RAG and custom LLMs
- Developed a Medical Cost Prediction Model that improved prediction accuracy and streamlined healthcare process recommendations
- Earned Oracle Generative AI and AWS Prompt Engineering certifications
I’m always eager to connect with professionals and teams on the leading edge of AI. Feel free to reach out for a casual coffee chat about any topic that piques your interest!
Experience
Vosyn
- Built and optimized end-to-end LLM pipelines and FastAPI microservices powering internal GenAI tooling. Reduced hallucination rates by 20% and improved model relevance through LoRA fine-tuning. Standardized prompt design practices and evaluation loops, accelerating R&D deployment velocity 3×.
Hoosier Community Network
- Developed React-based interfaces to test LLM chat flows in production and enhanced prompt consistency between backend APIs and user-facing components.
Infiniqe
- Developed responsive web interfaces using React.js, JavaScript, and SQL to enhance internal tools, and optimized team collaboration through structured Git workflows and code reviews.
Education
California State University - San Bernardino
GPA: 3.57 / 4.0
Gujarat Technological University, India
GPA: 3.62 / 4.00
Technical Skills
- Languages: Python, SQL
- ML & Data Science:PyTorch, TensorFlow, Scikit-learn, NLP, RAG, Fine-tuning
- Prompt Engineering: Zero/Few-Shot/chain-of-thought, hallucination detection, human-in-the-loop
- LLMs: GPT-4, Claude, Mistral, LLaMA, OpenAI
- Vector & SQL Databases: Chroma DB, FAISS, SQLite
- AI Tooling: LangChain, LangGraph, OpenAI API, OLLama, Groq, Unsloth, Cursor, TRAE
- Deployment & DevOps: Docker, Git/GitHub, MLflow, DVC, Streamlit , FAST API
- Cloud: AWS (EC2, S3, Lambda, Bedrock, SageMaker), GCP
- Other: Software Architecture, Agile (Scrum), Documentation
Projects
Medical Insurance Cost Predictor
A Python and PyTorch-based application that fine-tunes LLaMA 3.2-3B with LoRA for predicting medical insurance costs. Achieved 0.21 loss on 1,338 records while reducing training time by 75%, and delivered a full production pipeline with custom tokenization and inference.
Python PyTorch Transformers Unsloth Fine-tuning (LoRA) Pandas
Docubot-AI Doc Hub
A Python and Streamlit-based AI tool for real-time web data extraction using RAG and Chroma DB. Achieved 95%+ accurate, source-backed answers, improved document retrieval efficiency by 80%, and delivered a responsive, cloud-deployed UI.
Python Streamlit LangChain RAG Chroma DB Sentence Transformers
GenAI E-Commerce Chatbot
Built a LLaMA 3.3–powered chatbot with RAG and Chroma DB/SQLite for real-time product responses and 24/7 support, increasing user engagement by 40%. Reduced hallucinations by 30% through multi-prompt testing and deployed semantic routing with intent classification for accurate FAQ and product query handling via a Streamlit Cloud–hosted interface.
Python, Semantic Router, Chroma DB, SQLite, RAGPython RAG Gen AI Semantic Router Chroma DB SQLite
AI HR Management System
Developed a Python-based AI HR system automating employee lifecycle tasks, cutting manual workload by 60%. Leveraged Pydantic validation, FastMCP, and Gmail SMTP for reliable, enterprise-grade operations.
Python MCP SMTP Integration Claude Desktop
Gemstone Data End-to-End MLOps Pipeline
Created an end-to-end MLOps pipeline for gemstone classification using Python, MLflow, DVC, and AWS S3. Automated data preprocessing and model training, reduced data retrieval time by 30%, and ensured reproducible, scalable deployments via Docker and GitHub Actions.
Python MlFlow Airflow DVC GitHub Actions Aws S3 Docker
Certifications
Generative AI Professional
Oracle
Foundations of Prompt Engineering
AWS
Gen AI to Agentic AI with Projects
Codebasics
MLOps Certification
Ineuron
BUILD 2024 GEN AI Bootcamp
Snowflake