We build
ADAP Labs builds software, AI, and VR tools — and publishes research that pushes the boundaries of what technology can do.
Software, AI, and VR — built with research at the core.
Founded by University of Maryland researchers Arnav Dadarya and Anushk Pokharna, ADAP Labs builds software products, AI tools, and VR experiences — and publishes the research behind them.
AI Tools & Models
Building production-ready AI tools — from LLM frameworks to safety evaluation systems — used by developers and enterprises worldwide.
VR & Immersive Experiences
Designing VR and mixed-reality applications that blend intelligent interaction with immersive environments.
Software Engineering
Shipping reliable, scalable software — open source libraries, developer tooling, and full-stack products built to last.
Applied Research
Publishing peer-reviewed research at top venues, translating academic insights directly into working systems.
Enterprise Systems
Engineering AI infrastructure and data pipelines for Fortune 500 companies, handling millions of queries daily.
Products we've shipped and research we've published.
ImageBreak
An AI safety tool that stress-tests content moderation systems across leading AI models, identifying vulnerabilities before they reach production.
- Tests OpenAI, Gemini, and HuggingFace models
- Automated adversarial prompt generation
- Content moderation gap analysis
- Exportable safety reports
TS-DSPy
A TypeScript framework for building reliable AI-powered applications. Brings Stanford's DSPy methodology to the JS ecosystem with full type safety.
- Type-safe LLM signatures and validation
- ReAct reasoning and tool-use patterns
- Multi-provider support (OpenAI, Anthropic, Gemini)
- Built-in cost and usage tracking
Published on NPM: @ts-dspy/core
SynthoHive
An enterprise data tool that generates realistic synthetic datasets from complex relational schemas — privacy-safe and production-grade.
- Relational generation preserving foreign-key integrity
- Automated PII detection and anonymization
- Deep generative models (CTGAN, WGAN-GP)
- Scales to enterprise workloads via Apache Spark
The Founders
University of Maryland researchers bridging AI research with production-grade engineering.
ImageBreak: A Framework for Testing AI Model Safety
Authors: Arnav Dadarya, Anushk Pokharna
Arnav Dadarya
Co-Founder & Principal Researcher
B.S. Computer Science — University of Maryland
Published researcher. Founder of DefenX Security ($250K pre-seed). Building agentic AI at FedEx Dataworks, optimizing LLM serving with NVIDIA, and core contributor to Apache Kafka 4.0.
- Published Researcher — Preemptive Error Correction
- Founder & CEO — DefenX Security Solutions
- SWE Intern (Agentic AI) — FedEx Dataworks
- Apache Kafka 4.0 Core Contributor
Anushk Pokharna
Co-Founder & Software Engineer
B.S. Computer Science & Economics — University of Maryland
Advancing AI-powered experiences at Marriott International. Deep experience in DevOps modernization, biomolecular simulation research, and 12+ global hackathon wins.
- Software Engineer — Marriott International (AI)
- Research Assistant — UMD Clark School of Engineering
- World's Largest FinTech Hackathon Winner
- Co-Founded MetaJustice for social impact