agarwalshreshtha223@gmail.com
GitHubLinkedIn
Let's talk

Shreshtha
Agarwal

AI/ML Engineer — Mumbai, India

BuildingAIthatsurvivesoutsidenotebooks.

See my work
layer_01 / about

About me

I'm Shreshtha — an AI & Data Science undergrad at K.J. Somaiya School of Engineering (B.Tech, 2027), building AI that survives outside notebooks.

Professional bug creator by day, bug fixer by evening, AI engineer somewhere in between. Recently I've been deep in drug discovery and Spiking Neural Networks (SNNs) — and, obviously, plotting my next adventure.

EducationB.Tech in Artificial Intelligence & Data Science
K.J. Somaiya School of Engineering, Mumbai — expected 2027
CurrentlyAI/ML System Design Intern @ Biorad Medisys
Off the clockRSL London Acoustic Guitar — distinction & higher-achiever title

Languages

PythonC/C++SQLJavaScript

AI / ML

Deep LearningComputer VisionTransformersAdvanced RAGScikit-LearnNumPy / Pandas

Technologies

AWSDockerReactFlaskMongoDBMySQL
layer_02 / achievements
[0]Co-authored research on smart-contract DoS mitigation, published at IEEE ICBDS 2025.
[1]Logistics Head, Team Vision, KJ Somaiya (2025–26) — led operations for a technical fest with 500+ participants.
[2]Distinction & higher-achiever title, RSL London Acoustic Guitar grade debut (2024).
layer_03 / experience

Where I've worked

Jun 2026 — Present
AI/ML System Design Intern
Biorad Medisys · Medical devices
  • Built an AI-powered semantic search pipeline over FDA MAUDE medical-device reports, cutting adverse-event lookup from manual review to seconds across thousands of records.
  • Implemented a RAG architecture producing citation-grounded summaries traceable to source reports, improving auditability.
  • Shipped a natural-language search interface so non-technical users can query records without schema knowledge.
Jun 2025 — Aug 2025
Blockchain Security Intern
KJ Somaiya College of Engineering
  • Developed attack-resistant Ethereum auction smart contracts in Solidity, eliminating tested DoS vulnerabilities.
  • Reduced gas consumption by 12% while preserving security, lowering transaction costs for end users.
  • Published the resulting DoS-mitigation research at the IEEE ICBDS 2025 conference.
layer_04 / projects

Selected work

MAUDE Intelligence Platform

AI intelligence platform for semantic search and conversational analytics over the FDA MAUDE dataset — 35M+ medical-device adverse-event records. Includes an automated DFMEA/UFMEA/PFMEA generation pipeline that maps device specifications to FDA product codes and produces traceable, citation-grounded risk analyses.

Next.jsRAGLangGraphDocker

Ethereum Secure Auction System

Decentralized auction smart contract hardened against DoS attacks, with a secure withdrawal architecture that improves transaction reliability. The underlying DoS-mitigation research was published at IEEE ICBDS 2025.

SolidityWeb3Security

KA-GNN Paper Implementation

Implemented the Kolmogorov-Arnold Graph Neural Network (KA-GNN) paper end-to-end for molecular property prediction — a hands-on deep dive into drug-discovery pipelines, cheminformatics, and Kolmogorov-Arnold network architectures.

PythonPyTorchGNNRDKitCheminformaticsDrug Discovery
layer_05 / playground

Draw a digit

A real convolutional network, running entirely in your browser — draw a digit and watch it flow through the layers: pixels in, convolutions, pooling, prediction out.

loading model…
// the forward pass will appear here once you draw
output_layer / contact

Let's build
something.

agarwalshreshtha223@gmail.com