Hi, I'm Vansh Sheth

About

I am Vansh Sheth, a full-stack developer and AI enthusiast with experience building scalable web, mobile, and AI-driven systems. I am proficient in React Native, Next.js, Django, Python, WebSockets, Redis, PostgreSQL, and cloud deployments, and I actively use AI tools like Claude, v0, and Cursor to solve practical problems. I enjoy combining code, AI, and product thinking to create impactful solutions.

Skills

My Projects

Check out my latest work

I've worked on a variety of projects, from consumer apps to AI-driven platforms. Here are a couple of highlights.

Synthetic Data Distillation for Legal QA

Synthetic Data Distillation for Legal QA

Distilled a large Llama-3.1 legal QA model down to an 8B student using PEFT and quantization, achieving ~90% accuracy and strong NA detection while keeping inference cost 8–10× lower.

Python · PyTorch · HuggingFace · QLoRA · PEFT · vLLM

Multimodal Agentic RAG System

Built a hybrid BM25 + vector retrieval system with confidence control, async ingestion, and audio-aware chunking to serve multimodal queries in ~70–350ms over thousands of documents and embeddings.

LangChain · Milvus · Elastic · Redis

Flingle

A gamified campus dating app that combines real-time chat, smart matching, and XP-based progression to boost engagement. It focuses on creating a safe, interactive, and fun social experience for college students.

Aideit

AideitIQ is an AI-driven platform that converts large Reddit discussions into structured insights using LLMs, async processing, and data visualization, helping users quickly understand trends and sentiment at scale.

Education