I'm a B.Tech Computer Engineering student at Cummins College of Engineering for Women, Pune, with hands-on experience building AI/ML systems across LLMs, RAG, agentic AI, voice AI, NLP, and computer vision.

I enjoy moving between research and implementation: understanding a problem, building a system, investigating where it fails, and iterating from there.

Currently exploring the intersection of AI/ML, LLM systems, interpretability, reliability, and AI alignment.

B.Tech in Computer Engineering

Direct Second-Year Admission after completing a Diploma in Computer Engineering.

Diploma in Computer Engineering

97% Aggregate

• Large Language Model applications
• Retrieval-Augmented Generation (RAG)
• Agentic AI and multi-step workflows
• Prompt engineering
• Embeddings and vector search
• Conversational AI
• Multilingual AI systems

• AI alignment and safety
• Model interpretability
• AI reliability and robustness
• Representation learning
• LLM behavior and failure modes
• Responsible deployment of Generative AI

• OpenCV
• MediaPipe
• Hand tracking
• Gesture recognition
• Real-time human-computer interaction

• Speech-to-text
• Text-to-speech
• Voice agents
• Multilingual and Hinglish conversational systems
• Real-time WebSocket-based voice pipelines

A research-oriented AI system combining:

• Information retrieval
• NLP
• Transformer-based summarization
• Sentiment analysis
• Named-entity recognition
• Personalized recommendations
• Voice interaction

Explored approaches including BART, T5, Pegasus, content-based filtering, and collaborative filtering.

Built an LLM application that transforms website and document content into a searchable knowledge base for grounded conversational responses.

Focus areas:

• RAG pipelines
• Embeddings
• Vector search
• Context retrieval
• Agentic workflows
• Multilingual interaction
• Tool execution

Technologies:

LangChain LangGraph Groq Pinecone Supabase Redis Node.js

Built a conversational voice system connecting:

Speech → LLM → Response → Speech

Worked with:

• Whisper-based speech recognition
• Gemini TTS
• Sarvam AI
• WebSockets
• Node.js

Explored practical challenges involving latency, speech recognition, synthesis reliability, and long-running voice interactions.

Built real-time computer vision experiments using OpenCV and MediaPipe, including:

• Air Canvas
• Finger counting
• Hand tracking
• Gesture recognition
• Human-computer interaction experiments

Research examining the foundations, applications, limitations, and responsible deployment of Generative AI.

The work explores:

• Transformer-based LLMs
• Retrieval-Augmented Generation
• GenAI applications across industries
• Hallucination and reliability
• Bias and fairness
• Human oversight
• Responsible AI deployment

The paper also draws on firsthand experience building a multilingual RAG customer-service chatbot, Hinglish voice AI agent, and AI-powered idea validation platform.

Winnoviz Technologies

Working across:

• LLM application development
• RAG pipelines
• Agentic AI
• LangChain / LangGraph
• Multilingual conversational AI
• Voice AI
• Speech recognition and synthesis
• REST APIs
• WebSockets
• Vector databases
• AI system deployment

A major part of this experience has been understanding the gap between an AI system that works in a controlled environment and one that remains reliable in real-world use.

I'm increasingly interested in questions that go beyond simply making AI systems work.

When a model explains its own behavior, how do we know that explanation corresponds to the computation that actually produced the behavior?

How stable are learned behaviors when models encounter distribution shifts, adversarial prompts, or unfamiliar tasks?

Can AI systems reliably detect changes in their own behavior before those changes become visible through failures?

How should increasingly autonomous AI agents behave when objectives, context, tools, and instructions conflict?

These questions are currently shaping the direction of my AI/ML learning and research.

Python Java C C++ JavaScript SQL

LLMs NLP RAG Embeddings Agentic AI Computer Vision Speech AI

LangChain LangGraph Groq Gemini Whisper OpenCV MediaPipe

Node.js Express.js Flask REST APIs WebSockets

MySQL Supabase Pinecone Redis Docker

Vercel Render Netlify

• 📄 Research paper presented at IEI Pune
• 📄 Presented research work at PICT Impetus and Concepts '25
• 📄 Presented work at APCOER Technothon
• 🥉 3rd Place, Web Wizards HTML Competition
• 🎓 97% Aggregate, Diploma in Computer Engineering, Government Polytechnic Pune

• GitHub: Aarya-Chaudhari
• LinkedIn: Aarya Chaudhari
• Portfolio: Visit Portfolio
• Email: aaryachaudhari070@gmail.com