Research

Astyne Research Lab

Independent AI research advancing on-device intelligence, human-centric interfaces, and learning systems.

From what was said to what was meant. A small on-device model catches self-corrections and filler and trims them, so the prompt that goes on is shorter and simpler.

NSF Aligned Research

Our work advances critical priorities in artificial intelligence, focusing on deployable systems that operate under real-world constraints.

Research Focus

Astyne Research advances deployable artificial intelligence systems under real-world constraints—latency, compute, power, usability, and trust.

NSF Priority Alignment

  • Broadening AI deployment
  • Human-centered computing
  • Edge and embedded intelligence
  • STEM learning and accessibility
  • Societal benefit through technology

Broader Impacts

Enabling accessible, privacy-preserving AI for everyone

Privacy-Preserving AI

Local computation protects user data

Reduced Cloud Dependence

Edge processing minimizes infrastructure needs

Accessible Learning

Tools that work across all contexts

Energy Efficiency

Sustainable intelligent systems

Research Pillars

01

On-Device AI Systems

  • Quantized LLMs
  • Edge inference
  • Latency-critical pipelines
  • Privacy-preserving compute
02

Voice-First Interfaces

  • Speech-to-intent pipelines
  • Real-time correction
  • Conversational OS primitives
03

Learning & Cognitive Systems

  • Interactive lesson generation
  • Blackboard-style reasoning
  • Adaptive explanation depth
04

Human-Centered AI

  • Minimal cognitive load
  • Trust & transparency
  • Calm technology principles

AetherOngoing

Electrochemical HOCl Generation

Aether Home makes hypochlorous acid from water, salt and electricity. Our applied chemistry work asks how to make that solution consistently at home: how much free chlorine a batch carries, where its pH settles, and how electrodes and run time change both.

Free chlorine isn’t all HOCl

A free chlorine reading counts hypochlorous acid (HOCl) and hypochlorite (OCl⁻) together, and pH decides the split. Drag across the chart to see it.

97%HOCl
3%OCl⁻
pH 6.0
The textbook equilibrium at 25 °C (pKₐ ≈ 7.5), not Aether measurements. The shaded band is Aether’s design range, pH 5.0 to 6.5. Below about pH 4, dissolved chlorine gas starts to form as well, which this two-way split leaves out.

What we’re investigating

  • Free chlorine concentrationHow much free available chlorine each strength (Mild, Daily and Strong) produces, and how closely a batch lands on its target.
  • pH stabilityWhere a batch’s pH settles, and how far it drifts while the solution is kept.
  • Electrode materialsWhich electrode coatings hold up over many batches, and how they shift output and pH.
  • Electrolysis durationHow run time changes concentration, so each strength is reached by design rather than by trial.
  • RepeatabilityWhether the same water, salt and settings give the same batch, day after day and unit to unit.