Hi, I'm Shubham

MLShubham MaheshwariENGINEER

Shubham Maheshwari

M.Eng. Computer Science (UConn). I build and ship ML models, RAG pipelines, and data-driven products.

Let's Talk ●View Work
0.74
AUC credit model
176K
records modeled
96.6%
multi-modal acc.
<100ms
FastAPI inference
Machine Learning✦Data Science✦GenAI✦RAG Pipelines✦MLOps✦Deep Learning✦LLM Systems✦
Machine Learning✦Data Science✦GenAI✦RAG Pipelines✦MLOps✦Deep Learning✦LLM Systems✦
01

Projects

01AUC 0.7425

Credit Risk Scoring Engine

Full-Stack ML Platform

  • Trained LightGBM on 176K Lending Club records, reaching AUC 0.7425 vs a 0.7170 logistic baseline; GPU-accelerated via CUDA with SMOTE, validated with bootstrap CIs and a Wilcoxon signed-rank test across 5-fold CV.
  • Engineered an 8-class corporate credit-rating model (AAA–D, S&P-aligned) with 67.2% exact accuracy and 94.9% within-1-notch; integrated a Basel III stress-test simulator with per-request SHAP explainability over REST.
PythonLightGBMXGBoostSHAPFastAPIReact
Case study on request
0296.64% acc

Explainable Multi-Modal Deep Learning

Capstone

  • Designed a gated-fusion architecture reaching 96.64% accuracy, beating single-modality baselines by 14%.
  • Stress-tested under 10–30% noise with 18% higher accuracy retention; generated per-modality SHAP and Grad-CAM explanations on GCP Vertex AI.
PyTorchSHAPGrad-CAMMLflowGCP Vertex AI
Case study on request
0387% accuracy

Insurance Risk Advisor

Deployed GenAI + ML App

  • Built an end-to-end RAG pipeline using a ChromaDB vector store and sentence-transformer embeddings; extended it with a LangGraph agent that routes between vector search and web search; deployed on Hugging Face Spaces.
  • Trained a multi-class risk classifier on 50K+ records reaching 87% vs a 72% logistic baseline; deployed on AWS with real-time scoring, data-drift detection, and automated retraining triggers.
LangChainLangGraphLlama 3.1ChromaDBHuggingFaceStreamlit
Case study on request
02

Experience

Jul 2023 – Jul 2024

Software Developer Intern @ Sourcved Technologies

Ahmedabad, India

  • Integrated ML inference endpoints into REST-API-driven web apps; traced data flow from model backend to UI and identified 3 classes of serialization bugs causing silent data loss in prediction pipelines.
  • Cut UI defect rate by 40% over two release cycles via a pre-merge browser-matrix checklist (Chrome, Firefox, Edge), improving reliability of ML-powered feature rollouts.

Jan 2023 – May 2023

MERN Stack Developer Intern @ Arth Infosoft Pvt. Ltd.

Ahmedabad, India

  • Designed MongoDB schemas with optimized embedding vs referencing strategies for high-throughput workflows; built REST endpoints feeding structured data into analytical dashboards and React front ends.

Aug 2024 – May 2026

Student Manager @ University of Connecticut Dining Services

Storrs, CT

  • Managed scheduling, payroll, and onboarding for 40+ student employees; used data analytics to optimize shift allocation and cut scheduling conflicts by tracking attendance and peak-demand patterns.
03

Education

Aug 2024 – May 2026

University of Connecticut

M.Eng., Computer Science & Engineering (STEM)

Storrs, CT

GPA 3.36/4.00 · Machine Learning, Deep Learning, Data Mining, Cloud Computing

Jul 2019 – May 2023

Gujarat Technological University

B.E., Computer Science & Engineering

India

GPA 3.5/4.00

04

Stack

ML / AI

PyTorchTensorFlowScikit-learnXGBoostLightGBMSHAPGrad-CAMFeature EngineeringA/B TestingCredit Risk ModelingModel Monitoring & Drift Detection

GenAI / LLM

LangChainLangGraphRAG PipelinesChromaDBSentence TransformersGroq (Llama 3)HuggingFaceFine-tuning (LoRA / PEFT)

MLOps / Infra

MLflowFastAPIStreamlitDockerGitHub ActionsAWS (EC2, S3, SageMaker)GCP (Vertex AI, BigQuery)Databricks

Languages / Data

PythonSQLGitMySQLPostgreSQLMongoDBSnowflakePower BITableauReact

Certifications

✓Databricks — Generative AI Fundamentals
✓Databricks — Fundamentals Accreditation
✓AI Fluency: Framework & Foundations — Anthropic
✓Microsoft Azure AI Essentials: Workloads & ML on Azure