Dharmit Patel

Full Stack AI Developer

Master's student in Data Science at San Jose State University with expertise in building scalable systems, ML pipelines, and full-stack applications. Passionate about leveraging data-driven insights and cutting-edge technologies to solve complex problems.

Get In Touch Download Resume
def build_data_pipeline():
  "Scalable ML systems"
  data = ingest()
  features = engineer(data)
  model = train(features)
  return deploy(model)

Experience

Software Engineer
Techmicra IT Solutions • Ahmedabad, India
Feb 2023 – May 2024

• Built scalable Django web applications with advanced search, filtering, role-based authentication, and secure payment processing; optimized PostgreSQL indexes and queries to cut median response time by 40% and doubled concurrent throughput.

• Implemented CI/CD workflows with Jenkins to automate builds, tests, and deployments, reducing errors and accelerating release cycles.

• Performed systematic quality checks using automated analysis tools; tracked and resolved bugs efficiently using Jira, contributing to seamless collaboration in an Agile team.

• Executed structured test procedures and regression tests by following predefined test plans; documented results to support quality and compliance requirements.

Projects & Research

5-Agent Collaborative System Spec Dev Test Review Repair RAG: 127,264 CodeSearchNet Chunks via FAISS 93.9% Success Rate HumanEval 164 +25.9% Improvement vs Baseline 92.4 Quality Score Out of 100 3.6GB GPU Memory RTX 4090
🚀 Intelligent Agents for Software Development
LLMs, RAG, Multi-Agent, QLoRA Fine-tuning

Research Publication: Multi-agent collaborative system combining RAG and domain-specific fine-tuning. Integrated 5 specialized agents (SpecAnalyzer, Developer, Tester, Reviewer, Repair) with 127,264 CodeSearchNet code chunks indexed via FAISS for semantic retrieval.

Key Results:

  • ✓ 93.9% success rate on HumanEval (164 problems)
  • ✓ 75% first-try success rate (Phase 3)
  • ✓ 92.4/100 average quality score
  • ✓ 3.60 GB GPU memory with QLoRA (0.495% trainable params)
  • ✓ 1,500+ rigorous experiments with ablation studies
LLMs RAG Multi-Agent QLoRA DeepSeekCoder Code Generation FAISS HumanEval
Stock Forecasting: 99%+ Accuracy
Google Stock Quantitative Analysis
Python, TensorFlow, Scikit-learn

Production-style ML pipeline for financial time series achieving 99%+ forecast accuracy with LSTM and Random Forest models across 1,200+ trading days. Engineered portfolio optimization service with Sharpe ratio maximization and interactive dashboards.

TensorFlow LSTM Monte Carlo Pandas
UberEats Domino's Pizza Subway More Restaurants Your Cart Large Pizza $15.99 Garlic Bread $4.99 Checkout
Uber Eats Prototype
React.js, Node.js, MongoDB

Full-stack web application with Node.js backend and React.js frontend. Implemented user authentication, session handling, and REST APIs. Designed microservices architecture optimizing API response times by 30%.

React Node.js MongoDB JWT
RAG Medical AI Assistant
MediBot – AI Medical Chatbot
LangChain, HuggingFace, FAISS

AI-powered medical chatbot using Retrieval-Augmented Generation (RAG) with open-source tools. Processes medical documents, generates vector embeddings, and retrieves relevant information using LLMs.

LangChain RAG FAISS Streamlit
Positive Neutral Negative
Twitter Sentiment Analysis
Pandas, NLTK, Scikit-learn

NLP-based sentiment analysis model classifying tweets as positive, negative, or neutral, achieving 89.99% accuracy with Random Forest Classifier. Enhanced data-driven insights for marketing strategies.

NLP NLTK Random Forest Twitter API
Threat Suspicious Normal GNN + RAG Threat Detection
📊 Graph-based Few-shot Threat Detection Using RAG
Graph Neural Networks, RAG, Security, PyTorch

Master's Capstone Project (In Progress): Innovative approach combining Graph Neural Networks with Retrieval-Augmented Generation (RAG) for advanced threat detection with minimal training samples. The system leverages graph structure to model complex relationships in network data and augments GNN capabilities with retrieval-based context for improved anomaly identification.

Key Research Focus:

  • ✓ Few-shot learning paradigm for threat detection with limited labeled data
  • ✓ Graph neural network architectures (GCN, GAT, GraphSAGE) for node/edge classification
  • ✓ RAG integration for knowledge-augmented threat pattern matching
  • ✓ Network anomaly detection via graph representation learning
  • ✓ Scalable inference for real-time security monitoring
  • ✓ Cross-domain threat intelligence retrieval and adaptation
Graph Neural Networks RAG Few-Shot Learning PyTorch Geometric Network Security Anomaly Detection Graph Embeddings FAISS
🏆 AWARD-WINNING ACHIEVEMENT • April 2025
DWT Challenge Winner – Stanford LLM x Law Hackathon
Won the prestigious DWT Challenge by designing and deploying a lightning-fast AI system capable of accurately answering complex questions from lengthy legal documents with over 90% accuracy. This achievement demonstrates expertise in building production-grade AI systems that solve real-world problems while maintaining high accuracy standards.

Additional Recognitions

Code Java Challenge
Internshala Training's • Jan 2021
3rd Position + ₹10,000 Award • JavaFX Game Development
Web-Dev Battle
Internshala Training's • Jan 2021
3rd Position + ₹10,000 Award • Expense Manager in PHP

Skills & Expertise

Programming

Python, JavaScript, TypeScript, C/C++, SQL

Web Technologies

Django, React.js, Node.js, Express.js, HTML/CSS, ASP.NET

Databases

PostgreSQL, MySQL, MongoDB, Microsoft SQL Server

Data & ML

TensorFlow, Scikit-learn, Pandas, NumPy, Machine Learning, NLP

DevOps & Cloud

AWS, Docker, Kubernetes, Jenkins, CI/CD, AWS Glue

Data Visualization

Tableau, Power BI, Plotly, Matplotlib

Big Data

Kafka, Spark, Hadoop, Airflow, ETL, Snowflake

AI & GenAI

LangChain, LangGraph, Retrieval-Augmented Generation, LLMs

Education

Master of Science in Data Science
San Jose State University • San Jose, USA
Aug 2024 – May 2026
GPA: 3.63/4

Coursework: Big Data Technologies, Data Visualization & BI, Data Warehousing & Pipelines, Machine Learning, Probability & Statistics, Distributed Systems

Bachelor of Engineering in Information Technology
Gujarat Technological University • Ahmedabad, India
Aug 2019 – May 2023
GPA: 3.83/4

Coursework: Software Engineering, Data Structures & Algorithms, Artificial Intelligence, Computer Networks, Database Management, Operating Systems, OOP

Let's Connect

I'm always interested in hearing about new projects and opportunities. Feel free to reach out!

✉️ dharmitpatel98@gmail.com 📱 (408) 422-1304 🔗 LinkedIn 💻 GitHub