Shreya Mendi

MEng AI Student @ Duke | ML Engineer | AI Product Builder

|

About

About Shreya

Here is a little background

Focused on building intelligent systems that create real-world impact. I'm an MEng AI student at Duke University with experience in ML engineering, DevOps, and AI product development. I enjoy turning research ideas into practical solutions โ€” whether that's a safety-critical CV system for search and rescue, a bias audit for transportation AI, or a multimodal fashion discovery engine. I bring a strong problem-solving mindset and love working at the intersection of AI and social good.

โ˜• Fueled by chai tea, not coffee
๐Ÿ”ต Duke Blue through and through
๐Ÿค– Believes AI should be explainable
๐ŸŒฟ Building tech for social good
๐ŸŽ“ Duke University โ€” MEng in Artificial Intelligence

Experience

Duke Trust Lab

Research Assistant

Duke Trust Lab

PythonPyTorch

Jan 2026 โ€” Present

  • Co-authoring a NeurIPS submission on safe agentic AI behavior and intervention policy under faculty supervision; leading experimental design, evaluation framework construction, and manuscript preparation.
  • Developed benchmarking pipelines across 3+ research projects measuring model calibration, AI transparency, and decision reliability in high-stakes deployment settings.
BMW Group

Student Consultant (AI Capstone)

BMW Group

PythonFastAPI

Jan 2026 โ€” Present

  • Built and evaluated multiple ML regression models (LightGBM, Ridge, MLP, Optuna-tuned TabularMLP) on real BMW dealer sales data to predict inventory turnover speed; ranked product configurations by predicted sell performance across the full vehicle lineup.
  • Delivered explainable AI-driven inventory recommendations through a live REST API and interactive dealer dashboard, enabling business teams to make data-informed spec decisions with transparent model confidence.
Duke University

Teaching Assistant โ€” Managing AI in Business

Duke University

Jan 2026 โ€” Present

  • Facilitate discussion sections and office hours for a graduate-level course covering AI strategy, LLM deployment, and responsible AI adoption; supporting 40+ MEng students on project design and technical implementation.
  • Evaluate student AI system proposals and provide structured feedback on model selection, deployment tradeoffs, and business impact framing across industry-sponsored capstone projects.
Assetmantle

DevOps Engineer

Assetmantle

AWSKubernetesDocker

Sep 2023 โ€” May 2025

  • Reduced AWS/Hetzner infrastructure costs by 38% through architecture optimization and automated CI/CD (Docker, Kubernetes) with rollout/rollback policies, maintaining 99%+ uptime across distributed nodes.
  • Streamlined CI/CD pipelines with Kubernetes which increased deployment frequency and reduced time-to-market across the engineering org.
Hewlett Packard Enterprise

Software Development Intern

Hewlett Packard Enterprise

DockerJenkinsPythonLinux

Jan 2023 โ€” Jul 2023

  • Built Dockerized services on Linux, automating CI/CD workflows with Jenkins, shell scripting, and cloud computing platforms.
  • Integrated REST APIs in Python for monitoring using Grafana & Prometheus improving the overall stability of the product.

Skills

Hover over a skill for current proficiency

Python

95%

Python

PyTorch

85%

PyTorch

TensorFlow

82%

TensorFlow

FastAPI

88%

FastAPI

Docker

85%

Docker

Kubernetes

80%

Kubernetes

AWS

90%

AWS

Google Cloud

75%

Google Cloud

SQL

85%

SQL

Git

90%

Git

Jenkins

75%

Jenkins

Linux

85%

Linux

Pandas

90%

Pandas

NumPy

90%

NumPy

JavaScript

78%

JavaScript

TypeScript

72%

TypeScript

React

70%

React

Next.js

68%

Next.js

Jupyter

92%

Jupyter

MLflow

80%

MLflow

Projects

10 projects ยท hover to explore

When2Speak โ€” LLM Intervention Policy Agent
AI ResearchNLP

When2Speak โ€” LLM Intervention Policy Agent

Trained a lightweight RL policy network for multi-agent dialogue intervention using PyTorch and NLP. Reduced unnecessary interventions by 25% while maintaining task success rate, validated on a 10,000-dialogue simulation suite via A/B testing. Co-authoring a NeurIPS submission on safe agentic AI behavior under faculty supervision at Duke Trust Lab.

PythonPyTorchReinforcement Learning
UAV-SAR โ€” Aerial Human Detection for Search & Rescue
Computer VisionDeep Learning

UAV-SAR โ€” Aerial Human Detection for Search & Rescue

Fine-tuned Faster R-CNN on thermal SAR imagery with domain-specific augmentations (snow, smoke, sensor noise). Achieved 20% recall improvement under adverse conditions with <5% clean-data accuracy loss. Built end-to-end CV pipeline in PyTorch with custom data loaders and metrics optimized for safety-critical deployment.

PythonPyTorchTensorFlow
BMW Capstone โ€” Industrial AI Inventory Decision System
MLIndustry

BMW Capstone โ€” Industrial AI Inventory Decision System

Built and evaluated ML regression models (LightGBM, Ridge, MLP, Optuna-tuned TabularMLP) on real BMW dealer sales data to predict inventory turnover speed. Delivered explainable AI-driven inventory recommendations through a live REST API and interactive dealer dashboard, enabling data-informed spec decisions without requiring ML expertise.

PythonFastAPIDocker
Safe-T โ€” AI Equity Audit for Transportation Safety
AI EthicsData Analysis

Safe-T โ€” AI Equity Audit for Transportation Safety

Audited transportation AI allocation algorithms for racial and economic bias across Durham, NC โ€” documenting that Black residents (32% of population) account for 47% of pedestrian/cyclist crash victims, driven by systematic undercounting of demand in low-income and high-minority census tracts. Built an interactive Leaflet mapping platform comparing AI-predicted vs. need-based infrastructure allocation using Census, NCDOT crash records, and OpenStreetMap data.

PythonJupyterJavaScript
Mirror โ€” Reflective AI Mental Wellness Companion
RAGMental Health

Mirror โ€” Reflective AI Mental Wellness Companion

Built a reflective AI journaling companion using RAG over 8 psychological frameworks (CBT, IFS, NVC, Attachment Theory) to pattern-match emotional entries. Implemented cognitive distortion profiling across 4 distortion types and weekly self-awareness reports. Deployed full-stack Python/FastAPI on Railway with session persistence, mood trend tracking (1โ€“10 daily score), and a trigger map logging recurring emotional patterns.

PythonFastAPIDocker
CineStyle โ€” Multimodal Fashion Discovery from Film & TV
Multimodal AIRecommendation Systems

CineStyle โ€” Multimodal Fashion Discovery from Film & TV

Built a film-to-fashion identification platform using FashionCLIP (512-dim embeddings) and FAISS GPU vector search over 20,000 Fashionpedia garment crops across 46 categories. Four-stage recommendation pipeline (FAISS โ†’ NeuMF โ†’ SASRec โ†’ diversity filter) with measurable NDCG@10 gains over a popularity baseline. Deployed FastAPI on Railway, Next.js on Vercel; evaluated on Precision@K, Recall@K, and MAP@K with 500 synthetic users ร— 30 interactions.

PythonNext.jsFastAPIDocker
Inflationship โ€” Macroeconomic Forecasting Pipeline
Time SeriesForecasting

Inflationship โ€” Macroeconomic Forecasting Pipeline

Built a SARIMAX inflation forecasting pipeline combining port traffic indicators with CPI data, reducing forecast error to 0.67โ€“1.69% MAPE across major CPI categories. Validated model stability via 5-fold rolling cross-validation, achieving reliable predictive lift over CPI-only baselines.

PythonPandasNumPy
Alba โ€” AI Carbon Footprint Tracker
SustainabilityChrome Extension

Alba โ€” AI Carbon Footprint Tracker

Built a real-time, privacy-first Chrome extension estimating energy, carbon, and water footprint for AI prompts using model metadata and GitHub Models API. Implemented live footprint labels, heuristic + AI prompt optimization, and a client-side dashboard with daily impact summaries.

JavaScriptChrome ExtensionGitHub API
AI Audit โ€” EU AI Act Compliance System
AI ComplianceMLOps

AI Audit โ€” EU AI Act Compliance System

Built an EU AI Act compliance assessment system using TF-IDF + Logistic Regression, rule-based article evaluation (Articles 5, 6, 9, 10, 14), and automated remediation planning. Deployed full-stack ML workflow with MLflow, FastAPI, Docker, and Streamlit on Google Cloud Run, enabling explainable risk scoring and documentation audits.

PythonDockerFastAPIGoogle Cloud
Semantic Jury โ€” Legal Semantic Search Engine
NLPVector Search

Semantic Jury โ€” Legal Semantic Search Engine

Developed a semantic search engine for legal documents using embedding-based retrieval and citation-link modeling to surface semantically similar case law and statutes. Implemented vector search pipelines with passage ranking to support explainable legal research and faster knowledge discovery.

PythonTensorFlowJupyter

Contact

I've got just what you need. Let's talk.