Skip to content
View Dhyani2206's full-sized avatar
1 follower · 0 following

Highlights

  • Pro

Block or report Dhyani2206

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Dhyani2206/README.md

Dhyani Keyur Panchal

AI/ML Engineer | LLM Systems | Backend @ Scale

🚀 Building reliable, production-grade AI systems
🧠 Focused on LLMs, RAG pipelines, and real-world ML
⚙️ Bridging research → deployment with scalable systems

🧑‍💻 About Me

  • 🎓 Master’s in Computer Science @ Stevens Institute Of Technology (May 2026)
  • 🤖 Specialized in LLMs, RAG systems, and ML pipelines
  • ⚡ Experienced in building end-to-end AI systems (research → production)
  • 🧠 Focus: Reliability, scalability, and non-hallucinating AI systems
  • 🌍 Open to AI/ML Engineering roles (Full-time)

⚙️ Tech Stack

Languages:
Python | JavaScript | SQL

AI/ML:
PyTorch | Scikit-learn | XGBoost | LightGBM | PEFT | LoRA

LLM & RAG:
LLaMA | Mistral | FAISS | LangChain | Retrieval-Augmented Generation

Backend & Systems:
FastAPI | REST APIs | Microservices | Event-driven architecture

Tools & Infra:
Docker | Git | Linux | Uvicorn | ONNX Runtime

🚀 Featured Projects

🔹 Regulatory RAG System (LLM + FAISS + FastAPI)

  • Designed dual-retrieval pipeline for SEC filings + regulatory rules
  • Built deterministic answer engine with evidence grounding (zero hallucination focus)
  • Implemented semantic + structural query routing for financial documents

🔹 Fraud Detection System (ML + Behavioral Biometrics)

  • Developed hybrid model combining keystroke dynamics + phishing signals (~30 features)
  • Improved unsafe session detection by ~45%
  • Applied threshold tuning + imbalance handling for high-risk sensitivity

📫 Let's Connect

Python Machine Learning LLM FastAPI

Pinned Loading

  1. E-Store E-Store Public

    Basic application Made Using Visual Studio, Vb.NET and MySQL

    Visual Basic .NET 1

  2. Multiple-Timeseries-Forecasting Multiple-Timeseries-Forecasting Public

    Worked on timeseries dataset and implemented models that are statistical models v/s deeplearnign models to see the difference in their performance.

    Jupyter Notebook

  3. Search-Result-Enhancement Search-Result-Enhancement Public

    For boosting Search Algorithm, applied transformers and semantic similarity methods.

    Jupyter Notebook

  4. Domain_Specialized_LLaMA Domain_Specialized_LLaMA Public

    Fine-tuning LLaMA-3, Mistral-7B, and Phi-3 using QLoRA on a curated Data Science dataset (~8K high-signal Q&A pairs) to build a domain-specialized AI Data Science Tutor with semantic evaluation and…

    Jupyter Notebook

  5. Transformer-Based-Semantic-Retrieval-and-RAG-Optimization Transformer-Based-Semantic-Retrieval-and-RAG-Optimization Public

    Built and benchmarked a scalable semantic retrieval pipeline comparing lightweight bi-encoders and QLoRA-tuned large cross-encoders under real-world efficiency constraints

    Jupyter Notebook

  6. sec-regulatory-rag sec-regulatory-rag Public

    Deterministic evidence-first regulatory RAG assistant for SEC filings and rules, built with FastAPI and Streamlit.

    HTML 1 1