Services/Private AI

Private AI and voice transcription for healthcare, legal and other regulated teams

DevX Group LLC builds private AI for teams that cannot send sensitive data to someone else’s cloud: voice transcription and AI that run entirely on your own devices, like our Mac app Parlin. Clinics, hospitals and law firms can use our software as it is, or have our senior team integrate it into their own systems. Cloud AI is available when it fits.

Schedule a free consultationSee hourly pricing

Why private AI

Patient visits, client meetings and case notes are exactly the conversations people want transcribed and summarized, and exactly the data that should not leave the building. Most AI tools send the audio to a cloud service to process it. Ours do not have to.

Our speech recognition and AI models run on the computer or phone itself. The audio and the text stay on the device for that step, instead of being uploaded to be processed.

DevX Group LLC, San Diego

Three ways to use it

Start with the software as it ships, or have it built into the systems your team already uses.

  • Standalone: install Parlin on your team’s Macs for dictation into any app, meeting transcripts, summaries and action items.
  • Integrated: our team builds the same on-device transcription and AI into your own software, such as an EHR workflow, a case-management system or an internal tool.
  • Custom: private AI built for your data, from search across your own documents with a local model to tuning a model on your examples, running on hardware you control.

Security and compliance, stated plainly

No software is HIPAA compliant on its own; an organization is, through its policies, agreements and safeguards. What on-device processing changes is the data flow: when audio and text never leave your devices, there is no third-party processor to assess for that step.

We document exactly what is processed where, so your compliance or security team can review it. When a cloud model is the right tool, we agree the provider, the data handling and the retention with you before anything is built.

  • On-device speech recognition and local language models: no audio upload.
  • Data-flow documentation for your security and compliance review.
  • Least-privilege access, encrypted storage and audit logging in integrated builds.
  • Fast, secure cloud AI as an option, with providers and retention agreed in writing.

Who it is for

Doctors and clinics, hospital departments, therapists, law firms and any team whose conversations carry sensitive information. Parlin already lists doctors, lawyers and therapists among the people it is built for.

What we build with

Named tools, and the reason each one is on the list.

  • Local speech models on Apple silicon

    Speech-to-text that runs on the Mac or iPhone itself, fast and offline.

  • Ollama and local LLMs

    Summaries, drafting and Q&A from a model running on your own hardware.

  • Core ML and on-device embeddings

    Search and classification over your documents without uploading them.

  • Swift, SwiftUI and native integrations

    Mac and iPhone apps that fit into the tools your team already uses.

  • Claude and GPT models, when allowed

    Cloud models for heavier tasks, used only where you approve the data flow.

  • Next.js and Postgres

    The secure web dashboards and APIs around an integrated deployment.

How an engagement runs

Bi-weekly demos, billing per accepted milestone, then 30 days of stabilization and a 60-day warranty. Code and accounts stay in your name.

  1. 01Week 1

    Data-flow review

    We map what is recorded, where it is processed and who can see it, with your security or compliance lead in the room.

  2. 02Weeks 2 to 4

    Pilot

    Parlin or a working prototype on a few of your team’s machines, measured on your real vocabulary and workflow.

  3. 03Weeks 4 to 10

    Integrate

    On-device transcription and AI built into your systems, with access controls, logging and tests.

  4. 04Launch + 30 days

    Roll out and support

    Deployment to your team, training, then 30 days of direct support. Documentation and code are yours.

Proof, not promises

Shipped work that shows this service in practice.

Questions we get asked

Is your voice-to-text software HIPAA compliant?

HIPAA applies to organizations, not to software alone. Our transcription runs on the device, so audio and text are not sent to us or to a cloud service for that step. We provide data-flow documentation so your compliance team can confirm it fits your policies.

Can you integrate private transcription into our existing system?

Yes. Our senior team builds the same on-device speech recognition and AI into your own software, such as an EHR or case-management workflow, with access controls and logging, after a data-flow review with your team.

Can we just use your app without a custom project?

Yes. Parlin is available for Mac today and Parlin Recorder for iPhone. Many teams start there and only commission an integration once they know what they need.

Do you also build cloud AI?

Yes. When the data and your policies allow it, cloud models are faster to build with and scale easily. We agree the provider, data handling and retention with you in writing before building.

Who else is this for besides healthcare?

Law firms, therapists, financial advisors, HR teams and anyone whose conversations contain information they would rather not upload. The same private approach works for all of them.

Talk to the engineer directly

A 30-minute call with Max Sheikhizadeh: what you need, what it costs, and whether we are the right fit.

Schedule a free consultationSend a project brief
How we work: ownership, milestone billing, warranty and references