Ehretz GeskinHe/They
Head Research Engineer, InthraOS
Edge AILocal LLMsPrivacyOn-device ML
"Data-driven research that builds and trains SLMs/LLMs for the edge, verifies compliance end-to-end, and delivers models you can trust to run safely at scale.
Ehretz leads InthraOS research on how we design, build, and train small and large language models—and how they run safely on devices at scale. The work centers on privacy, compliance, and AI security, ensuring our systems meet real-world rules and protect people by default. "
Research Focus
- On-device inference & quantized SLMs
- Privacy Receipts & runtime evidence
- AI control plane, RBAC, staged rollouts
Open to
- Academic collaborations
- Community research pilots
- Standards & policy workstreams
Featured writing
Article
The Future of AI Is Local: Why Sovereign, On-Device Intelligence Wins
Latency, privacy, and reliability from device-first design.
Edge AIPrivacyLatency
Read →
Article
Governing AI Like an Enterprise System: The Control Plane You’re Missing
Models, prompts, and indices as managed, attested assets.
MLOpsControl PlaneRBAC
Read →
Professional Experience
Head Research Engineer
InthraOS
Aug 2025 – Present · New York (Hybrid)
- Lead research across privacy-preserving systems, edge inference, and control plane design.
- Collaborate with external labs; translate findings into productized capabilities.
Laboratory Technician
Columbia University
- Collected/analyzed experimental data on drosophila escape behaviors using MATLAB; presented results.
Research Assistant
Baruch College
- Studying dissociation in transgender and nonbinary adults; scholarly research & data handling.
Why InthraOS
Deep bench in experimental research and applied engineering, proven rigor in sensitive, high-stakes domains, and a builder’s bias to convert research into shippable, evidence-producing systems.
Collaborate with InthraOS Research
Interested in joint studies, benchmarking, or standards work? Let’s talk.