Shruti Karmarkar
Hello, I'm

Shruti Karmarkar

Data Scientist · Machine Learning · AI Engineering
View My Work Let's Connect
A note from me

I'm a curious person by nature, I've always had a "figure it out" mindset, and data science gave that instinct a real outlet.

That matters more right now than ever. AI is moving faster than anything I've seen, and staying useful means adapting fast, which I'm actively doing.

But speed isn't the real skill. The real instinct today is asking why: why this model, why this tool, is it actually valuable or just hype? That means finding the right metrics, understanding how a model actually behaves, and forming a real hypothesis instead of taking a shiny new tool at face value. That's what keeps me motivated right now.

What I Do

Three disciplines.
One direction.

01

AI Agents & LLM Systems

Production AI applications, from multi-agent pipelines to LLM copilots, that automate complex, high-stakes workflows.

See projects → 02

Data Intelligence

Turning raw data into the story a business doesn't know it's telling, with models that scale.

See projects → 03

Analytics & Storytelling

Bringing quantitative rigor to decisions that used to run on instinct alone. Data as a creative tool, not just a report.

See projects →
Selected Work

Projects I'm excited to show you.

Real problems, shipped systems, and the numbers to back them up. Click in.

01 AI Agents & LLM Systems

01

ClaimIQ

Multi-agent insurance adjudication: autonomous agents that read, reason over, and resolve claims end-to-end.

Next.js 14TypeScriptClaude APILangSmith
View Live Site →
02

SaaS Churn Copilot

A LangGraph churn-intelligence platform that predicts at-risk accounts and recommends the save.

PythonXGBoostLangGraphFAISSPostgreSQL
View Live App →

02 Data Intelligence

03

HCC Coding Gap Classifier

A fine-tuned small language model that flags Medicare risk-adjustment documentation gaps. It matched a frontier LLM on accuracy (88.9% vs 75.9%) at a fraction of the cost.

PythonLoRAMLXPhi-3.5-miniClaude API
View on GitHub →
04

Neutrogena Sales Forecasting

Review sentiment + topic modeling fused with a Random Forest to forecast sales. R² = 0.94.

VADERLDARandom ForestPython
View on GitHub →

03 Analytics & Storytelling

Macy's Competitor Intelligence Dashboard05

Macy's Competitor Intelligence

An interactive Tableau dashboard built on real SEC & Statista data, mapping Macy's against the field.

TableauSEC DataStatista
View Dashboard →
Bank of America Financial Intelligence Dashboard06

Bank of America Financial Intelligence

A financial-intelligence Tableau dashboard distilling performance signals into a single executive view.

TableauFinancial Modeling
View Dashboard →
Experience

The road so far.

Summer 2025
Data Science Intern, AI Sourcing
Xometry

Built data-driven sourcing intelligence for a manufacturing marketplace, applying ML to match supply with demand at scale.

Jan 2023 - Jun 2024
Research Data Scientist
Gilbert Research Center

Led applied research turning experimental datasets into models and insight, bridging the gap between raw findings and decisions.

Sep 2024 - May 2026
M.S. Computer Engineering
NYU Tandon School of Engineering · AI / Data Science

Specializing in AI and data science: building multi-agent systems, ML pipelines, and the analytical foundation for the work I want to do next.

Jul 2022 - Dec 2022
Data Analyst, Part-Time Co-op
AI Adventures

Built predictive models and analytics solutions at AI Adventures across multiple client industries, including retail and sports media.

The Toolkit

What I build with.

AI & LLMs

  • LangChain
  • LangGraph
  • RAG Pipelines
  • Multi-Agent Systems
  • Claude API
  • FAISS
  • LangSmith

ML & Modeling

  • XGBoost
  • Random Forest
  • NLP / VADER
  • LDA
  • Feature Engineering
  • A/B Testing

Engineering

  • Python
  • TypeScript / Next.js
  • PostgreSQL
  • Docker
  • PySpark

Analytics & Viz

  • Tableau
  • SQL
  • Pandas
  • Matplotlib
  • Statistical Modeling
  • Data Storytelling
Where I'm Headed

I don't want a seat at the table.
I want to redesign what's on it.

I'm drawn to the problems no one has cracked yet. If that's the kind of work you're doing, let's talk.

Say Hello

Let's build something
that matters.

If you're working on something ambitious, I want to hear about it.

spk9869@nyu.edu LinkedIn GitHub Resume