smridhi@portfolio ~ bash
Smridhi Gupta speaking on LLM Security
about.md

Who I Am

I started in a research lab, curious about how machines learn. Now I build the systems that put that learning to work — at scale, in production, for real people.

Technology Analyst at Citibank, working on AI security platforms, MCP tooling for 2,500+ daily users, and LLM evaluation pipelines. My work lives at the intersection of research-grade thinking and production-grade engineering.

From transformer-based medical AI research at IGDTUW → ML research at the AI Institute of South Carolina → agentic workflows at Citi. Along the way: mentoring students through Intel × Digital India's AI4Youth, speaking at security and developer meetups, and earning a Research Excellence Award from the Delhi Education Minister.

AI's future is shaped by people who translate complexity into curiosity.

2500+daily users impacted
40%triage time reduced
94.3%model accuracy
4+talks given
gallery.sh --moments

In The Room

YUVAi mentoring

Intel × Digital India · AI4Youth · Ahmedabad

Mentoring the Next Gen of AI Builders

Guided grades 9–12 students through hands-on ML model implementation at the YUVAi Rapid Modelling Workshop.

LLM Security talk

AI + Security Community Meetup · Bengaluru

LLM Security ≠ Traditional Security

Why semantic AI threats demand an entirely different mental model from conventional security.

"You can't write a regex that catches all prompt injections."

— AI + Security Meetup, Bengaluru 2024

JS Pune speaker

JS Pune Meetup · Feb 2026 · Globant

Rethinking JS Testing with AI

AI-augmented testing workflows for JavaScript engineers.

Citi Red Carpet

Citibank · Red Carpet Rewards

Recognised for Engineering Ownership & Impact

Citi RedCarpet Award for delivery impact across AI security and agentic platform projects.

Research Award

IGDTUW Foundation Day · Delhi

Research Excellence Award · ₹50,000 Grant

Awarded by Shri Ashish Sood, Hon'ble Education Minister of Delhi.

More stages,
more stories
incoming.

Talks · Panels · Workshops · Podcasts

work.log

Where I've Built

Citibank

Jul 2024 – Present

Pune, India

Technology Analyst II (Promoted from Analyst I)

  • AI Security Screening Platform — automated restricted-entity checks & escalation workflows for internal compliance.
  • MCP Tooling — Java-based tools for 2,500+ daily users, structured entity extraction (ticker, ISIN, BloombergCode) for tool-using AI workflows.
  • Prompt & Model Evaluation Harness — regression & recall pipelines enabling safe AI rollout in production.
  • FIX-AI Platform — agentic platform with guardrails for 1,100+ employees, cutting manual triage time by 40%.
  • High-Throughput Batch & Caching Pipelines for large-scale list-check scans supporting downstream AI services.

AI Institute of SC

Jan 2024 – Jul 2024

Columbia, USA (Remote)

ML Research Intern

  • Built reproducible ML pipelines for benchmarking and system evaluation of research models.
  • Contributed to evaluation methodology across multiple research systems.

IGDTUW

Mar 2022 – Jul 2024

Delhi, India

ML Research Intern

  • Transformer-based medical AI achieving 94.3% accuracy on lung cancer classification.
  • Published 4 papers; work recognised with Delhi Education Minister's Research Excellence Award + ₹50k grant.
papers.bib

Research Papers

2024

IEEE CONECCT 2024

Transforming Lung Cancer Diagnosis with Twin Vision Transformers

Proposes a Twin Transformer (TwinViT) architecture for CT scan classification into NSCLC subtypes — Squamous Cell, Adenocarcinoma, Large Cell Carcinoma — achieving 94.3% accuracy and 94.49% precision.

Vision TransformersMedical AILung CancerIEEE
View Paper ↗

2024

Conference Paper · April 2024

Enhanced Efficiency in Lung Cancer Classification via Deep Learning Ensembles

Explores ensemble deep learning architectures for improved lung cancer classification efficiency and accuracy on CT imaging datasets, improving diagnostic performance over single-model baselines.

Deep LearningEnsemble MethodsCT Imaging
View on ResearchGate ↗

2023

Conference Paper · March 2023

Pulmonary Nodules Binary Classification using CNN and LSTM

Proposes a CNN + LSTM hybrid architecture for binary classification of pulmonary nodules on LIDC-IDRI and LUNA16 datasets, achieving 86.89% accuracy. Cited 1 time.

CNNLSTMPulmonary NodulesMedical Imaging
View on ResearchGate ↗

2022

Conference Paper · December 2022

Binary Classification of Pulmonary Nodules using Long Short-Term Memory (LSTM)

Implements LSTM-based binary classification of pulmonary nodules for early-stage lung cancer detection. Cited 4 times — the most cited paper in the series.

LSTMBinary ClassificationEarly Detection
View on ResearchGate ↗

2024

Systematic Review · July 2024

Prospect of LLMs and NLP for Lung Cancer Diagnosis: A Systematic Review

A systematic review of the emerging role of Large Language Models and NLP techniques in lung cancer diagnosis workflows — bridging Smridhi's medical AI and LLM expertise.

LLMsNLPSystematic ReviewLung Cancer
View on ResearchGate ↗

2023

Conference Paper

Fine-Tuned Framework With Transfer Learning for Pulmonary Nodule Detection

Fine-tuning pre-trained CNN architectures via transfer learning for pulmonary nodule detection, comparing multiple model variants and color spaces for improved detection on standard CT datasets.

Transfer LearningFine-tuningNodule Detection
View on ResearchGate ↗
skills.json

What I Know

Languages

PythonJava TypeScriptJavaScript C++GoSQL

AI & LLM

LLM IntegrationAgentic Workflows MCP ToolingModel Evaluation GuardrailsPrompt Engineering Transformers

Backend & Cloud

MicroservicesKafka FastAPISpring AWSGCP DockerKubernetes

Data & Observability

PostgreSQLMongoDB RedisLogging MonitoringPerf Debugging
awards.log

Recognition & Wins

Citi Red Carpet

Citibank

Red Carpet Rewards & Recognition

Recognised for engineering ownership and delivery impact across AI security and agentic platform projects.

Research Award Delhi

IGDTUW · Delhi Education Ministry

Research Excellence Award · ₹50,000 Grant

Presented by Shri Ashish Sood, Hon'ble Education Minister of Delhi, for commendable undergraduate research in medical AI.

  • Top 15 / 3,000+ teamsMercedes-Benz TechSign Challenge
  • Mentor — Intel × Digital IndiaAI4Youth Program
  • Technical Head, IEEE IGDTUWLed 15+ events · 250+ participants
  • Promoted to Analyst IICitibank, within first year

open --collab smridhi-gupta

Let's build something
interesting.

Working on something at the edge of AI? Let's connect.

Connect on LinkedIn