Aryan Thakar

Secure full-stack engineer with applied AI experience.

Computer Science, University of Maryland — expected Dec 2026

About

I build reliable, secure, user-focused software. My work spans full-stack engineering, applied AI, and web security across industry internships, NSF-funded research, and teaching secure web development at the University of Maryland.

Experience

  1. AI Security Intern @ MITRE

    – McLean, VA

    • Built an experimental pipeline for testing robustness of Open Policy Agents using attack cases generated with Promptfoo, reducing successful attacks by 30% for output-rail and tool-authentication components.
    • Increased throughput and usability of AI vulnerability analysis software by developing a Docker-containerized React/FastAPI web application using Redis for batch queueing and Celery workers for concurrent processing; identified parser-validation vulnerabilities through testing varied system prompts.
    • Won the AI Learning Track Hackathon by collaborating to fine-tune a phishing-link classification model that achieved the competition’s lowest log-loss score.

  2. SEO Tech Developer

    – Remote

    • Built a movie/TV review application using Flask, Jinja, and SQLAlchemy to prompt thoughtful, structured reviews.

  3. Teaching Assistant — CMSC388J: Building Secure Web Apps

    University of Maryland

    • Assisted instruction in secure, full-stack web apps.
    • Topics and tooling included Python/Flask, Django, MongoDB, and modern front-ends such as Svelte/React.

  4. Creative Software Developer Intern @ RedPeg Marketing

    – Alexandria, VA

    • Improved a retrieval-augmented generation chatbot’s query resolution quality by 30%.
    • Built a CustomGPT workflow with Microsoft Graph API + OpenAI function calling, saving 30–40 min/week per employee.

  5. Research Fellow @ NSF TRAILS (UMD)

    – College Park, MD

    • Designed inclusive interview protocols to reduce bias in automated video interviewing systems.
    • Implemented feature extraction with dlib/OpenCV, improving accuracy by 0.1 MSE.

Projects

DocuWell

OpenAI • React • MongoDB • Auth0

Patient–doctor chatbot prototype; improved interaction quality by 20% with structured follow-ups.

ReactOpenAIMongoDB

View code →

Bias in Public Datasets

Python • Fairlearn • LIME

Fairness analysis on U.S. Census & UCI Heart Disease models; surfaced parity/accuracy trade-offs and gender feature over-weighting.

PythonFairlearnLIME

View code →

Authentic Screen Talk

Flask • Jinja • SQLAlchemy

Movie/TV review site that prompts thoughtful reviews beyond star ratings.

FlaskJinjaSQLAlchemy

View code →

Cons App — Ethical Consumption

React Native • Android Studio

Prototype that helps users find ethical product alternatives, reducing search time by ~50%.

React NativeMobile

View project →

SEO Dev — Review App

Flask • Jinja • SQLAlchemy

App used in SEO Tech Developer program to elicit higher-quality opinions with prompt-guided form flows.

FlaskSQLAlchemy

Skills

JavaPythonCRustOCaml JavaScriptSQLReactReact Native Node.jsFlaskMongoDB TensorFlowPyTorchOpenCV FairlearnLIME