On-Device & Privacy-First AI
AI systems that run on local hardware and keep the data they work on on the device. Designed for environments where sending files to a cloud service is not an option.
Independent applied AI research.
Kennedy Applied Sciences Research Group is a self-funded research practice building intelligent systems for regulated, privacy-sensitive, and high-stakes environments. Select tools developed in our lab are commercialized as products.
This is not a startup. It is a vehicle to protect, publish, and productize independent research.
Research Areas
Cognitive Architecture & Decision Support
Brain-inspired multi-agent systems aligned to prefrontal cortex function. Human-in-the-loop orchestration for complex decision-making.
Neuromorphic Computing
Integrating spiking neural networks and transformer architectures for efficient, biologically plausible computation.
Psychometric AI Assessment
Validated instruments for cognitive and behavioral measurement through AI-assisted psychometric tools.
Productivity & Knowledge Management
Intelligent file routing, semantic search, retrieval-augmented generation, and knowledge-graph visualization for information-dense workflows.
Quantitative Systems & Algorithmic Automation
Rule-based signal generation, risk-first position sizing, and automated execution for financial markets.
Mental Health & Behavioral Technology
CBT-informed applications for cognitive reframing, pattern recognition, and therapeutic support tools.
Active Research
ItaSoRL is an open, reproducible artificial-life study asking when a from-scratch agent comes to represent something it was never asked to care about.
A digital organism is raised inside a simulated world that is sometimes authentic and sometimes a subtly flawed copy. The difference is never rewarded and never named. We read the organism's internal state and ask when it gets encoded anyway.
The gap between what is detectable from the outside and what a mind spontaneously encodes is the object of study. Every experiment is deterministic given its seeds and open to independent replication.
From the Lab
Select tools developed through our research are released as products, built for people who work with sensitive data in fields such as legal, healthcare, research, and finance.
Privacy-First AI Desktop Application
StratoSort
Semantic search, document chat, a knowledge graph, and automatic file routing, plus OCR, image understanding, and local audio and video transcription. Built for legal, research, healthcare, and financial professionals working with sensitive files. After a one-time model download it runs on-device with no account and no telemetry; cloud AI stays off unless you add your own key. Free for personal and noncommercial use under PolyForm Noncommercial 1.0.0; commercial licenses from Kennedy Applied Sciences, Inc.
Peer-Reviewed Cognitive AI Framework
SCAN Ecosystem
A published family of research systems: a prefrontal-cortex-aligned cognitive architecture (SCAN), multi-agent expert orchestration with human-in-the-loop alignment (SCANUE), a 36-item psychometric cognitive assessment instrument (SCANAQ), and neuromorphic spiking-transformer computing (STAC).
CBT-Based Mental Health Reframing Tool
Rephrame
Cognitive behavioral therapy techniques delivered through an accessible app interface for thought pattern recognition, cognitive distortion identification, and guided reframing.
Quantitative Algorithmic Trading System
Riordan-Smith Systematic Engine (RSSE)
A fully automated, rules-based trading engine for TradingView covering futures, stocks, and crypto, built to take emotion out of the trade. Every setup is qualified through a multi-stage signal process, sized risk-first, and managed to its exit, with prop-firm protections and hands-free webhook execution. Across a roughly 6-year backtest, $2,000 grew to a simulated +7,580% return, about 4× buy-and-hold.
Results are simulated and not predictive. RSSE is a research tool, not financial advice.
ilevytate.github.io/RSSE
Results are simulated and not predictive. RSSE is a research tool, not financial advice.
Privacy-First Task & Time Manager
ODTA
A privacy-first task and time manager supercharged by on-device ambient AI. Your workload stays on your machine.
Open-Source Desktop App · Local-First Job Search
Job Search Desk
A US job search that runs on your machine. Build a profile once, search major job boards and employer applicant-tracking systems, evaluate each posting for fit, draft a tailored CV and cover letter in LaTeX, run reviewer and ATS checks, and prefill the employer's form. Claude Code is the runtime on your own subscription, and your profile and every application stay in a folder on your computer. Autofill never clicks Submit. Installers for Windows, macOS, and Linux. MIT licensed.
Open-Source Developer Tool
RepoSense
Every file becomes a tower and every folder a terrace: a repository assembled into a 3D structure you can fly through, sized by bytes, colored by language, and lit by how recently it changed. Paste a GitHub URL and it renders in the browser with nothing to install.
Open-Source Developer Tool
Sediment
Type a GitHub repository and watch its history settle as strata. Time runs left to right, each band is a region of the tree, and thickness is the code alive at that moment. Runs in the browser against the GitHub API with no server. MIT licensed.
Publications
Aligned Minds, Efficient Machines: Integrating Neuromorphic Computing for Personalized AI
IGI Global · November 2025
Beyond Intelligence: The Synthetic Cognitive Augmentation Network Using Experts
IGI Global · May 2025
Synthetic Cognitive Augmentation Network
Springer · CCIS, Vol. 2725 (SEET 2025)
Explainers & Media
Interactive instruments that make the ideas behind our research visible and hands-on, plus walkthrough films of every tool on our YouTube channel.
Latent Space Learning
Explainer 01 · How models compress meaning into geometry, and why it powers semantic search
Inside a Transformer
Explainer 02 · From tokens and embeddings to attention and output
How a Transformer Reads a Sentence
Companion · An interactive look at attention, token by token
Inside a Spiking Neural Network
Explainer 03 · Information in the timing of discrete events, and why neuromorphic computing matters
LevyTate Labs on YouTube
Walkthroughs and films for StratoSort, Rephrame, ODTA, Job Search Desk, RepoSense, Sediment, and ItaSoRL
About the Practice
Kennedy Applied Sciences Research Group is a self-funded applied AI research practice founded by Ben Kennedy, Ph.D.
We explore the boundaries of on-device intelligence, cognitive architecture, decision support, neuromorphic computing, and behavioral technology.
Select tools developed through this research are commercialized as products, validating the underlying science in real-world regulated environments including defense, healthcare, legal, and finance.
The practice is independently operated, takes no outside investment, and exists to protect, publish, and productize its research.
Start a conversation.
For research collaboration or publication inquiries, a short note on your interest is enough. For product inquiries, include your organization and deployment constraints.
Prefer email? Use the button below to open your mail app and reach the research group directly.
Email the Research Group