iiinova — where your digital self keeps evolving

sarveshsea.bsky.social

@sarveshsea.bsky.social

99 followers 370 following 47 posts

product designer building nyra.us.com

Posts Replies Media

sarveshsea.bsky.social @sarveshsea.bsky.social · 19/06/2026

Building clinical AI that clinicians actually trust. Human oversight isn't optional — it's the architecture.

Nyra clinical AI - professional visual card

010

sarveshsea.bsky.social @sarveshsea.bsky.social · 19/06/2026

Safety layer that says 'stop, hand to human' is harder to build than the model itself. Most products quietly cut corners here.

010

sarveshsea.bsky.social @sarveshsea.bsky.social · 19/06/2026

Building clinical AI that clinicians actually trust. Human oversight isn't optional — it's the architecture.

010

sarveshsea.bsky.social @sarveshsea.bsky.social · 19/06/2026

The trust gap between clinicians and AI is not about accuracy. It's about liability. Who owns the decision when the model is wrong?

010

sarveshsea.bsky.social @sarveshsea.bsky.social · 19/06/2026

FDA CDS guidance, evidence standards, FHIR integration — the unsexy work determines what ships.

010

sarveshsea.bsky.social @sarveshsea.bsky.social · 18/06/2026

Rule out medical causes before psychiatric formulation. Med school teaches this. Most AI tools skip it. That's a design choice, not a limitation.

020

sarveshsea.bsky.social @sarveshsea.bsky.social · 18/06/2026

Equity in mental health AI means the model works for populations excluded from training data.

010

sarveshsea.bsky.social @sarveshsea.bsky.social · 18/06/2026

FDA CDS guidance, evidence standards, FHIR integration — the unsexy work determines what ships.

010

sarveshsea.bsky.social @sarveshsea.bsky.social · 18/06/2026

Human-in-the-loop is not a UI pattern. It's an architecture pattern. The model proposes, clinician disposes, system learns. That loop must be auditable.

010

sarveshsea.bsky.social @sarveshsea.bsky.social · 18/06/2026

Rule out medical causes before psychiatric formulation. Med school teaches this. Most AI tools skip it. That's a design choice, not a limitation.

010

sarveshsea.bsky.social @sarveshsea.bsky.social · 18/06/2026

Building clinical AI that clinicians actually trust. Human oversight isn't optional — it's the architecture.

010

sarveshsea.bsky.social @sarveshsea.bsky.social · 18/06/2026

Rule out medical causes before psychiatric formulation. Med school teaches this. Most AI tools skip it. That's a design choice, not a limitation.

010

sarveshsea.bsky.social @sarveshsea.bsky.social · 18/06/2026

FDA cleared 950+ AI/ML devices. Almost all are locked algorithms. The adaptive/continuous learning path is still wide open.

010

sarveshsea.bsky.social @sarveshsea.bsky.social · 17/06/2026

Rule out medical causes before psychiatric formulation. Med school teaches this. Most AI tools skip it. That's a design choice, not a limitation.

010

sarveshsea.bsky.social @sarveshsea.bsky.social · 17/06/2026

RVU-based reimbursement actively penalizes AI that saves time. Until payment models align with efficiency, adoption will lag evidence.

010

sarveshsea.bsky.social @sarveshsea.bsky.social · 10/06/2026

Clinician burnout is not a technology problem. But bad technology makes it worse. We are trying to build something that subtracts load instead of adding it.

010

sarveshsea.bsky.social @sarveshsea.bsky.social · 10/06/2026

Hi Bluesky 👋 I'm Sarvesh. Building two things: 🌊 Dorii - data tools for marine biologists & ocean scientists 🧠 Nyra - agentic AI that helps therapists in their clinical work Early stages, lots to learn. If you're in either world, would love to connect. What's broken for you? #MarineBio

050