AI-Native · Agent-Driven · R&D Innovation
Where Deep Research
Meets
Operational Certainty
This is where the work lives.
Published research, filed patents, and a team that
has been building mission AI since before it had a name.
has been building mission AI since before it had a name.
0 +
U.S. PATENTS
0 +
Years of R&D
CVPR / NeurIPS
Published At
FOCUS AREAS
Research That DrivesMission Readiness
Our work spans the technical foundations that make agent-native AI reliable, explainable, and deployable in the most demanding operational environments.
Agentic AI & Orchestration
Multi-agent systems reason, coordinate, and act across complex operational workflows — without requiring manual configuration or data team support.
Multimodal Retrieval & Grounding
Cross-source retrieval architectures ground AI predictions in real evidence — from imagery and geospatial data to logistics records and maintenance histories.
Predictive Readiness Modeling
Probabilistic models surface degradation signals early — converting fragmented maintenance, personnel, and supply chain data into forward-looking readiness estimates.
Explainable AI & Drill-Down Transparency
Interpretability research makes AI recommendations auditable — so operators can interrogate a recommendation, understand the rationale, and decide with confidence.
Edge & DDIL Deployment
Architectures are optimized for degraded, disconnected, intermittent, and limited environments — bringing AI decision support to the operational edge.
Geospatial & Visual Intelligence
Vision-language models and geospatial retrieval systems extract operational intelligence from imagery — supporting ISR, logistics, and situational awareness.
Research & Intellectual Property
Where Research Becomes Protected Advantage
Our science and our IP are the same effort viewed from two angles — peer-reviewed publications establish the methods, and our patent portfolio protects how they're applied in the field.
Published Research
Featured · Computer Vision · Geospatial AI
Pinpoint: Grounded Worldwide Image Geolocation via Cross-Source Retrieval and Reranking
A retrieve-and-rerank pipeline grounds image geolocation predictions in geotagged visual data. Achieving state-of-the-art accuracy across IM2GPS3k, YFCC4k, and OSV-5M benchmarks at 0.098 seconds per image deployable in classified or air-gapped environments without external MLLM APIs.
Nika Chuzhoy, Brian Hu, Amit A. Arora, Jae Ro, Sarthak S. Sahu
Nika Chuzhoy, Brian Hu, Amit A. Arora, Jae Ro, Sarthak S. Sahu
Read Paper ↗
Clinical AI · Epidemiology
Influenza Activity and Regional Mortality for Non-Small Cell Lung Cancer
A population-level analysis linking CDC influenza-like illness activity with SEER non-small cell lung cancer mortality across 13 U.S. states — finding 24% higher mortality during high-flu months among 202,485 patients. Supports influenza vaccination and prevention as potentially important strategies for reducing cancer-associated mortality.
Scientific Reports (Nature), Vol. 13, Article 21674 · Kinslow et al. incl. Donalek & Amori
Scientific Reports (Nature), Vol. 13, Article 21674 · Kinslow et al. incl. Donalek & Amori
Nature ↗
Human-Computer Interaction · Data Visualization
Immersive and Collaborative Data Visualization Using Virtual Reality Platforms
An immersive virtual-reality platform for exploring and collaboratively analyzing high-dimensional scientific datasets — using 3D visualization to improve perception of data geometry, pattern discovery, and intuitive understanding beyond traditional desktop interfaces.
IEEE International Conference on Big Data, p. 609 · Donalek, Djorgovski, Davidoff et al.
IEEE International Conference on Big Data, p. 609 · Donalek, Djorgovski, Davidoff et al.
arXiv ↗
Patent Portfolio
Category 01
Network Characterization & Device Intelligence
Methods provide network risk scoring, key terrain identification, device criticality analysis, and automated device type prediction and identification.
Category 02
3D Data Visualization & Network Extraction
Systems render and navigate high-dimensional 3D data and extract network structures from complex datasets.
Category 03
Explainability & Natural Language Querying
Methods provide network explainability and natural language interfaces for querying complex operational data.
Category 04
ML-Based Scenario Simulation
Software engines detect high-impact scenarios using machine learning-based simulation.
Selected Patents
Patent No.
Title
Granted
Status
US 10,454,597
Systems and Methods for Locating Telecommunication Cell Sites
Oct 2019
granted
US 11,928,123
Systems and Methods for Network Explainability
Mar 2024
granted
US 12,086,134
Systems and Methods for Natural Language Querying
Sep 2024
granted
US 12,174,729
Systems with Software Engines Configured for Detection of High Impact Scenarios with Machine Learning-Based Simulation
Dec 2024
granted
US 12,223,570
Systems and Methods for High Dimensional 3D Data Visualization
Feb 2025
granted
US 12,244,635
Computer-Based Systems Configured for Network Characterization and Management Based on Risk Score Analysis
Mar 2025
granted
US 12,301,613
Computer-Based Systems Configured for Network Characterization and Management Based on Automated Key Terrain Identification
May 2025
granted
US 12,536,202
Systems and Methods Configured for Dataset Sampling
Jan 2026
granted
US 18,484,373
Systems and Methods for 3D Data Visualization and Network Extraction
Jul 2026
granted
US 19,205,793
Computer-Based Systems Configured for Network Characterization and Management Based on Device Criticality Score
Jul 2026
granted
THE TEAM
Scientists, Strategists, and Service members
Our research team combines deep technical expertise with real operational experience — building AI that works in the field, not just in the lab.
Dr. Ciro Donalek
CTO & Co-Founder
Former Computational Staff Scientist at Caltech with 100+ publications in Nature, Neural Networks, and IEEE Big Data. Holds foundational patents behind the Virtualitics platform and led the iViz project at Caltech, the precursor to Virtualitics Explore.
Sarthak Sahu
Chief AI Officer
B.S. in Computer Science from Caltech, where he conducted machine learning research with Professor Yisong Yue. Previously built classified defense software at Raytheon and sports analytics models at Second Spectrum that outperformed published benchmarks.
Aakash Indurkhya
Chief Product Officer
Caltech Computer Science graduate who has shaped the Iris platform for a decade, bridging AI, analytics, and mission outcomes. Founded and taught a big data frameworks course at Caltech and previously consulted on AI across infrastructure, retail, and sports analytics.
Justin Gantenberg
Chief Engineer
B.S. in Computer Science with a specialization in Computer Graphics from UIUC. Has directed Virtualitics' engineering and infrastructure since 2016, drawing on prior experience building VR launch titles for Oculus Rift and is co-author on multiple platform patents.
JOIN THE TEAM
Build AI That Operates
Where It Matters
We're hiring researchers and engineers who want their work to have a measurable operational impact. If you want to build AI that supports decisions in high-stakes environments, we'd like to talk to you.
View Open Roles →