I'm an AI Engineer with a B.E. in Artificial Intelligence & Data Science (Mumbai University, 2025), focused on building production-ready Generative AI and Agentic AI systems.

I work across the full AI stack — from designing RAG pipelines and multi-agent workflows with LangChain, LlamaIndex & CrewAI, to building AI automation with n8n and OpenAI/Gemini APIs. I'm passionate about taking ideas from rapid POC to modular, deployment-ready backends.

class VaishnavThorwat:
    def __init__(self):
        self.role         = "AI Engineer | GenAI & Agentic AI"
        self.education    = "B.E. in AI & Data Science '25 — Mumbai University"
        self.location     = "Navi Mumbai, Maharashtra, India"
        self.focus        = ["LLM Orchestration", "RAG Pipelines", "Agentic Workflows",
                             "NLP", "Deep Learning", "AI Automation"]

    def current_stack(self):
        return {
            "GenAI":       ["LangChain", "LlamaIndex", "CrewAI", "OpenAI API", "Gemini 2.0"],
            "Automation":  ["n8n", "Streamlit", "Webhook Pipelines"],
            "ML/DL":       ["TensorFlow", "Keras", "Scikit-learn", "CNN", "Bi-LSTM"],
            "Languages":   ["Python", "SQL"],
        }

    def say_hi(self):
        print("Let's build AI that actually does something useful!")

• 🤖 Deepening expertise in Agentic AI — multi-agent orchestration and tool-use patterns
• 🔗 Exploring LLM fine-tuning and advanced RAG optimization techniques
• 🌐 Building more production-ready AI backends with FastAPI + LangChain
• 📊 Experimenting with AI-powered analytics automation using n8n + Gemini

Stack: LlamaIndex · LangChain · Google Gemini 2.0 Flash · Vector Embeddings

An intelligent document querying system with a dual-index RAG architecture. Dynamically routes queries between semantic search and tree summarization based on intent — with persistent on-disk vector indexing.

• 📌 Key Highlights: Dual-index architecture, dynamic tool-routing, QuestionsAnsweredExtractor metadata enrichment
• 🎯 Impact: Reduced redundant LLM API calls by ~40% via shared StorageContext

Stack: CrewAI · Gemini 2.0 Flash · Streamlit · SerperDevTool · OWASP

A 3-agent agentic workflow (Developer, Security Engineer & Tech Lead) that autonomously reviews pull requests for code quality and security vulnerabilities across 4 severity levels.

• 📌 Key Highlights: Agentic tool-routing, real-time OWASP scanning, structured JSON output, auto-blocking logic
• 🎯 Impact: Shipped as both a CLI tool and interactive Streamlit web app with one-click report download

Stack: CNN-Bi-LSTM · TensorFlow · Keras · NLP Pipeline

A hybrid deep learning model for real-time cyberbullying detection on Hinglish text, with a complete NLP preprocessing pipeline and client-server inference framework.

• 📌 Key Highlights: Hinglish multilingual text classification, tokenization, stopword removal, sequence padding on 21K+ entries
• 🎯 Impact: 91% validation accuracy · 91.45% precision · 90.65% recall

Stack: Python · Scikit-learn · Lasso & Ridge Regression · Feature Engineering

Applied advanced feature engineering across 128 variables to predict house prices — used Lasso and Ridge regression to isolate the top 20 most impactful features, reducing model dimensionality while improving interpretability.

• 📌 Key Highlights: 128→20 feature reduction, Lasso/Ridge regularization, hierarchical clustering, EDA on 1,400+ properties
• 🎯 Impact: R² score of 0.81 — identified overall quality and living area as strongest price drivers

B.E. — Artificial Intelligence & Data Science Engineering Terna Engineering College, Navi Mumbai | Mumbai University | 2021 – 2025

Relevant Coursework: Machine Learning, Deep Learning, Natural Language Processing, Data Mining, Database Management, Probability & Statistics, Data Structures & Algorithms

I'm actively seeking AI Engineer / GenAI Engineer / Applied AI roles where I can build real, production-grade intelligent systems.

• 💼 LinkedIn
• 📧 thorwatvaishnav@gmail.com
• 🌐 Portfolio: Coming Soon!