MS Computer Science • AI/ML · Full-Stack · iOS · Android

Building AI · Web · Mobile systems with a research mindset.

Thesis on deepfake/OOD detection (IEEE FG 2026). I build ML pipelines, full-stack web apps, and native iOS & Android apps — shipping production-ready products end-to-end.

View Projects LinkedIn GitHub Resume

Publications

Projects

Years Experience

638K+

Records Analyzed

Scroll

About Me

I’m Mili Patel — building AI systems that don’t break when the world changes, and shipping full-stack products that make those systems usable.

Research-first engineering

I treat deepfake detection as anomaly detection — not just classification.

My thesis focuses on detecting deepfake facial images under real-world distribution shifts (new manipulations, unseen datasets, changing lighting, compression artifacts, and model drift). That means I care about robustness, calibration, and decision boundaries — not just accuracy on a single benchmark.

Alongside research, I build production-grade systems: clean UI, fast APIs, evaluation dashboards, and reproducible pipelines. I like work where machine learning meets engineering — the part where results become tools.

AI/ML + Vision

OOD detection, deepfakes, calibration, pipelines.

Full-Stack

Next.js apps, APIs, dashboards, deployments.

Anomaly DetectionDeepfake ForensicsOOD / Distribution ShiftExperiment TrackingDashboardsProduction ML

What I’m building now

Robust deepfake anomaly detection pipeline

Training on clean splits, validating with controlled OOD sets, and testing against unseen manipulation methods — with score analysis + dashboarding for model behavior.

Research

Thesis Track

Full-Stack

Product Mindset

ML + CV

Primary Focus

UNT

MS CS

Background

MS Computer Science @ University of North Texas (Thesis).
Teaching / mentoring mindset — I enjoy making complex topics clear.
I build projects that can be demoed, measured, and shipped.

My principle

If it can’t be evaluated, monitored, and explained — it isn’t done.

Education

Academic foundation in AI, machine learning, and systems.

Graduated

Master of Science

Computer Science

University of North Texas

Denton, TX, USA

2024 – 2026

GPA: 3.625 / 4.0 · Distinguished Student of the Year 2024/25
Thesis: DF-OOD: Real-Only Deepfake Detection via Confidence Dynamics Under Perturbations (Defended Apr 2026)

RELEVANT COURSEWORK

Machine LearningDeep LearningComputer VisionData MiningNLPAdvanced Algorithms

Completed

Bachelor of Technology

Computer Science & Engineering

Institute of Technology, Nirma University

Ahmedabad, India

2020 – 2024

GPA: 3.14 / 4.0
Coursework: Data Structures, Machine Learning, Computer Vision

RELEVANT COURSEWORK

Data StructuresAlgorithmsDBMSOperating SystemsComputer NetworksAI/ML Foundations

Projects Showcase

Full-stack builds, ML research, dashboards, and edge projects — all in one place.

Flagship (Thesis)

Anomaly Detection for Deepfake Facial Images (Master’s Thesis)

University of North Texas • 2024 – 2026 (Expected)

Research-driven anomaly/OOD detection for deepfake facial images under distribution shift — beyond closed-set classification.

Focus on distribution shift and generalization across manipulation types.
Build reproducible pipeline + robust evaluation + demo-ready web app.
Analyze model behavior using OOD signals and calibration-style insights.

Research pipeline + demo-ready app

PythonPyTorch/TensorFlowComputer VisionOOD/AnomalyNext.jsDashboards

GitHub

AI/ML • Full-Stack • Mobile

ResearchMind AI — Research Intelligence Platform (Web & iOS)

Personal Project • 2026

8-feature AI research platform with retrieval-augmented cross-paper Q&A, inline citations, and long-context (128K-token) streaming inference — plus a native iOS app for paper discovery and AI-assisted reading.

8 AI features • Web + iOS • 128K context • IEEE PDF + LaTeX

Live Demo GitHub

Built an 8-feature AI research platform: Paper Vault, Cross-Paper Q&A with inline citations, Literature Review Generator, Research Gap Analyzer, Paper Comparison Matrix, Hypothesis Forge, IEEE Paper Forge (PDF + LaTeX export), and live paper discovery.
Integrated ArXiv and Semantic Scholar APIs for live academic paper search with citation counts and open-access PDF links; streamed Groq/Llama 3.3 70B (128K context) responses in real time.
Developed a native iOS application in Swift and SwiftUI for research-paper discovery, reading, and AI-assisted Q&A using MVVM architecture and URLSession async networking.
IEEE Paper Forge generates complete review papers in IEEE format with selectable sections, page targets, and citation count — exports as PDF and IEEEtran LaTeX.

Next.jsTypeScriptSwiftSwiftUIMVVMURLSessionGroq / Llama 3.3Semantic Scholar APIArXiv APITailwind CSS

AI/ML • Full-Stack • Mobile

CareerCopilot AI — Full-Stack AI Platform (Web & Android)

Personal Project • 2026

Full-stack AI career platform of 6 tools with real-time streaming and live job search across 50+ countries — plus a native Android app for resume analysis and job recommendations.

Live • 6 AI tools • Web + Android • 50+ countries

Live Demo GitHub

Architected and shipped solo a full-stack web platform of 6 AI career tools: Resume X-Ray with ATS scoring, Job-Match Engine, Mock Interviewer, Career Roadmap Builder, Cover Letter Forge, and Smart Job Board (Adzuna API across 50+ countries); used by real users across multiple countries.
Built a native Android application in Kotlin/Jetpack Compose (MVVM) that consumed the backend APIs for resume analysis and job recommendations, including file-upload functionality with Coroutines and Retrofit.
Streamed Groq + Llama 3.3 70B responses in real-time using ReadableStream and Next.js App Router; deployed on Vercel with server-side PDF extraction.

Next.jsTypeScriptKotlinJetpack ComposeMVVMCoroutinesRetrofitGroq / Llama 3.3Adzuna APITailwind CSS

Mobile • AI/ML

Ledgr — AI-Assisted Personal Finance App (iOS & Android)

Personal Project • 2026

The same AI-powered personal finance app built natively on both platforms — iOS in Swift/SwiftUI and Android in Kotlin/Jetpack Compose — with an offline-first architecture and an LLM assistant for spending insights.

Native iOS + Android • Offline-first • LLM-powered spending assistant

Built the same AI-assisted personal-finance app natively on iOS (Swift, SwiftUI, SwiftData, Swift Charts) and Android (Kotlin, Jetpack Compose, Room, Compose Charts) on an MVVM architecture with offline-first local persistence.
Integrated Claude/OpenAI API through typed networking layers (async/await + URLSession on iOS, Coroutines + Retrofit on Android) for a natural-language assistant that categorizes transactions and answers spending questions.
Visualized spending patterns with Swift Charts / Compose with secure on-device key storage (Keychain / encrypted storage), reusable components, and full light/dark mode support.

SwiftSwiftUISwiftDataSwift ChartsKotlinJetpack ComposeRoomMVVMCoroutinesRetrofitClaude APIOpenAI API

AI/ML • Full-Stack

FinAgent — Agentic GenAI Assistant on Java/Spring Boot + RAG

Personal Project • 2026

Enterprise-grade agentic AI assistant for finance-document Q&A — Java/Spring Boot microservice, Python/FastAPI inference, RAG over PostgreSQL/pgvector, multi-step LangGraph flows, and an Angular frontend.

Enterprise microservices • RAG + pgvector • LangGraph agentic flows • Docker CI/CD

Built an agentic AI assistant for finance-document Q&A: a Java/Spring Boot microservice exposes secure REST endpoints and orchestrates a Python/FastAPI inference service, with an Angular/TypeScript front end — mirroring an enterprise service architecture.
Implemented retrieval-augmented generation over PostgreSQL with pgvector embeddings for grounded, cited answers; designed multi-step agentic flows (LangGraph) with tool use for lookup, summarization, and structured extraction.
Containerized all services with Docker and a CI/CD pipeline (GitHub Actions), plus responsible-AI guardrails (groundedness and hallucination checks) on every response.

JavaSpring BootPythonFastAPILangChainLangGraphClaude APIOpenAI APIPostgreSQLpgvectorDockerGitHub ActionsAngularTypeScript

Full-Stack

Sarthak Group Tuition Website

Sarthak Group Tuition • 2025

Comprehensive tuition management platform for students and tutors with online learning features.

Full-featured tuition platform

Live Demo

Developed full-stack tuition management website with student portal, course management, and scheduling.
Implemented online class booking, payment integration, and progress tracking features.
Built responsive UI with Next.js and integrated backend APIs for real-time updates.

Next.jsReactTypeScriptTailwind CSSNode.jsMongoDB

Full-Stack

Smart Inventory Management System

University of North Texas • Apr 2025 – May 2025

Real-time retail dashboard with alerts, email notifications, and exports.

Real-time dashboard + automated alerts

GitHub

Full-stack solution (Streamlit + MySQL) with real-time dashboards and low-stock alerts.
Designed normalized schemas with indexes/triggers; role-based access control.
Enabled exports to Excel/PDF and email notifications.

PythonStreamlitMySQLSQLEmail NotificationsData Modeling

Data/BI

AI-Driven Crime Pattern Analysis

University of North Texas • Mar 2025 – Apr 2025

Large-scale crime analytics with dashboards for hotspots, trends, and outcomes.

638K+ records analyzed

GitHub

Analyzed 638,000+ U.S. crime records; built Tableau/Power BI dashboards.
Cleaned + modeled data in Python to extract state/weapon/demographic insights.
Delivered visual insights for final presentation and decision-making.

PythonData CleaningPower BITableauAnalyticsVisualization

Edge/Hardware

Smart Attendance System

University of North Texas • Jan 2025 – Apr 2025

Raspberry Pi + OpenCV pipeline for real-time face recognition and attendance automation.

Real-time recognition on edge

GitHub

Built real-time facial recognition pipeline using Raspberry Pi + OpenCV.
Processed live classroom feeds; optimized for low latency.
Added engagement-style metrics to enhance reporting.

Raspberry PiOpenCVPythonComputer VisionEdge Deployment

Edge/Hardware

AI-Based Object Detection on FPGA (Funded Project – ISRO)

Ahmedabad (Funded Project) • Aug 2023 – Dec 2023

YOLOv3 inference optimization on PYNQ FPGA for lower latency object detection.

30% lower latency • 40% better sync

GitHub

Implemented YOLOv3 inference on PYNQ FPGA, reducing end-to-end latency by 30%.
Designed Raspberry Pi + motor-controller network; improved synchronization by 40%.
Optimized hardware–software integration for real-time performance.

YOLOv3PYNQ FPGARaspberry PiPython/C++Edge AIEmbedded

Hackathon

MineD Hackathon — Journal Rejection Predictor

Nirma University • Mar 2023

ML model to flag likely journal rejections + Streamlit prototype for formatting issues.

35% fewer manual revisions

GitHub

Built classifier (Random Forest / Logistic Regression) to flag likely rejections.
Reduced manual revisions by ~35% via early rejection prediction.
Shipped Streamlit prototype highlighting formatting issues in real time.

PythonScikit-learnRandom ForestLogRegStreamlit

Hackathon

HackInfinity — Real-Time Captioning & Translation (EdTech)

DA-IICT, Gandhinagar • Feb 2023

Captioning tool for accessibility with live speech-to-text + multilingual translation.

Accessibility-first EdTech demo

GitHub

Designed accessibility-focused captioning for hearing/language impairments.
Integrated speech-to-text + multilingual translation for live support.
Used OpenCV + TensorFlow components to support real-time flow.

OpenCVTensorFlowSpeech-to-TextTranslationPython

Skills & Expertise

A curated view of the tools I use to build ML systems and full-stack products.

Generative AI & LLMs

RAG, agents, fine-tuning, and multi-provider LLM apps

Advanced

RAG Systems

Advanced

Prompt Engineering

Advanced

Agentic Flows (LangGraph)

Advanced

LangChain / LlamaIndex

Advanced

Fine-tuning (LoRA / QLoRA)

Advanced

MCP (Model Context Protocol)

Advanced

Multi-provider LLM APIs

Advanced

LLM Evaluation

Advanced

Responsible / Ethical AI

Advanced

Machine Learning & AI

Deep learning, anomaly/OOD detection, GPU training, MLOps

Advanced

PyTorch

Advanced

TensorFlow / Transformers

Advanced

Scikit-learn / XGBoost

Advanced

Computer Vision

Advanced

OOD / Anomaly Detection

Advanced

GPU DDP + FP16 training

Advanced

Hugging Face

Advanced

MLOps

Intermediate

Mobile — iOS & Android

Native iOS (Swift/SwiftUI) and Android (Kotlin/Jetpack Compose)

Advanced

Swift

Advanced

SwiftUI

Advanced

MVVM + async/await

Advanced

SwiftData / CoreData

Advanced

URLSession / Networking

Advanced

Kotlin

Advanced

Jetpack Compose

Advanced

Coroutines + Retrofit

Advanced

Room / Android Arch

Advanced

Xcode / Android Studio

Advanced

Backend & Microservices

Spring Boot, FastAPI, Node.js, REST APIs, microservices

Advanced

Java / Spring Boot

Advanced

Spring MVC / Spring Cloud

Advanced

FastAPI

Advanced

Node.js

Advanced

Flask / Django

Intermediate

REST APIs

Expert

Microservices Architecture

Advanced

Web / Frontend

Next.js, React, Angular, streaming SSR

Advanced

Next.js (App Router, SSR, streaming)

Advanced

React

Advanced

Angular / TypeScript

Advanced

Tailwind CSS

Advanced

Framer Motion

Advanced

Responsive UI

Advanced

Programming Languages

Across ML, systems, mobile, and full-stack

Advanced

Python

Expert

Java

Advanced

TypeScript / JavaScript

Advanced

Swift

Advanced

Kotlin

Advanced

SQL

Advanced

C++

Intermediate

Bash

Intermediate

Data & Vector Databases

Relational, NoSQL, and vector stores for AI

Advanced

PostgreSQL + pgvector

Advanced

FAISS / ChromaDB

Advanced

MySQL

Advanced

MongoDB

Intermediate

Pandas / NumPy

Advanced

Data Science & Analytics

Feature engineering, dashboards, and insights

Advanced

Feature Engineering

Advanced

SQL Analytics

Advanced

Power BI

Intermediate

Tableau

Intermediate

Splunk (basics)

Intermediate

Cloud & DevOps

AWS, Docker, Kubernetes, CI/CD

Intermediate

AWS (Certified AI Practitioner)

Advanced

OpenShift

Intermediate

Docker

Intermediate

Kubernetes (basics)

Intermediate

GitHub Actions (CI/CD)

Intermediate

Linux

Advanced

Testing / Automation

QA, scripting, and CI pipelines

Intermediate

Pytest

Intermediate

Selenium

Intermediate

Automation scripting

Intermediate

Tools / Platforms

Daily engineering toolkit

Advanced

Git / GitHub

Advanced

Xcode / Android Studio

Advanced

VS Code / Jupyter

Advanced

AWS Kiro (Spec-Driven Dev)

Intermediate

AI Tools & Development Practices

Using AI as force multipliers

Advanced

Claude / OpenAI / Groq APIs

Advanced

GitHub Copilot / Claude Code

Advanced

Responsible AI Guardrails

Advanced

Clean Code & Scalability

Advanced

Prototype & Iterate

Advanced

Experience Journey

A timeline of my roles — teaching, internships, leadership, and hands-on engineering work.

Teaching Assistant / Instructional Assistant — CSCE 4010 (Social Issues in Computing)

University of North Texas

Denton, TX Spring 2026 — Present

Assist course delivery under Dr. Moawia Eldow (CSCE 4010).
Support students through guidance, clarifying concepts, and feedback.
Help with grading, rubrics, and maintaining consistent evaluation.

TeachingMentoringGradingCS Education

Teaching Assistant / Instructional Assistant — CSCE 4010 (Social Issues in Computing)

University of North Texas

Denton, TX Fall 2025

Graded assignments/discussion work and supported course operations.
Helped students navigate course requirements and improved submissions via feedback.
Contributed to smooth course logistics and timely communication.

TeachingEvaluationCommunication

Summer Camp Computing Instructor

University of North Texas

Denton, TX May 2025 – Jun 2025

Taught middle & high school students programming fundamentals using Python and embedded systems.
Designed hands-on lessons touching AI, robotics, and hardware programming.
Worked with staff to deliver a smooth 6-week learning experience.

PythonRoboticsEmbeddedInstruction

Research & Development Intern

Space Applications Centre (SAC), ISRO

Ahmedabad, India Jan 2024 – May 2024

Developed a GUI for a Stewart-platform telescope to capture high-resolution galaxy images.
Designed a network topology using Raspberry Pi + motor controllers; improved synchronization by ~40%.
Built automation scripts for motor control to strengthen hardware–software integration.

Raspberry PiAutomationGUISystems

Software Development Intern (Remote)

Vas Ventures Pvt. Ltd.

Remote Jun 2023 – Jul 2023

Built a Tkinter GUI to automate bulk WhatsApp messaging; reduced manual effort by ~80%.
Applied ML models in CallAstro to predict marriage status using web-scraped data (~85% accuracy).
Automated testing with Selenium; reduced bug-detection time by ~60%.

TkinterSeleniumMLAutomation

Vice President — ACES (Association of Computer Engineering Students)

Nirma University

Ahmedabad, India 2023 – 2024

Led ACES activities and coordinated student teams to deliver technical events and workshops.
Collaborated with faculty and peers to execute department-level programs and student engagement initiatives.
Mentored junior students and supported community-driven learning and collaboration.

LeadershipCommunityEvent ManagementMentoring

View LinkedIn post

Research & Publications

Peer-reviewed work and research contributions — plus my ongoing thesis work in deepfake anomaly detection.

Master’s Thesis — Deepfake Anomaly Detection

Detecting deepfake facial images under distribution shift using anomaly/OOD signals — with robust evaluation and behavior analysis.

Defended Apr 2026

Goal

Robust deepfake detection beyond closed-set classification.

Method

OOD scoring, calibration analysis, and distribution-shift evaluation.

Output

Reproducible pipeline + metrics dashboard + demo-ready app.

PublishedIEEE

DF-OOD: Real-Only Deepfake Detection via Confidence Dynamics under Perturbations

IEEE 20th Int'l Conf. on Automatic Face and Gesture Recognition (FG 2026)

2026

View Paper

Authors: Mili Patel, Ajita Rattani

Novel OOD detection framework trained exclusively on real images — no deepfake data needed at training time.
Achieved AUROC 99.12% on DF40 and 95.53% average across FaceForensics++ via energy-based scoring with ODIN perturbation.
Improved generalization to unseen deepfake generators by reframing detection as one-class learning.

PublishedSpringerDOI: 10.1007/s11831-025-10298-5

Advancements and Challenges in the Use of Artificial Intelligence for Coronary Artery Disease Diagnosis: An Integrated Review

Archives of Computational Methods in Engineering (Springer)

June 2025

View Paper

Authors: Heni Mehta, Mili Patel, Manav Vakharia, Parita Oza

Comprehensive review of AI and deep learning methods for coronary artery disease diagnosis.
Analyzed challenges including data quality, generalization, interpretability, and clinical adoption.
Discussed future directions for reliable AI-assisted clinical decision systems.