Skip to content
Adi's Digital Garden

Deep dive · 01

The problem — the long gap between doctor visits

For someone managing a chronic condition, the hardest part isn't the appointment. It's the weeks and months after it — adhering to a plan, decoding instructions, and handling small daily decisions alone.

Where PHP lived

Doctor
visit
weeks–months of daily friction — adherence, "what do I do now?", missed-dose questions — where PHP is the bedside assistant
Next
visit
The clinician sets the plan; PHP helps the patient actually live it between visits.

Concretely, that meant helping patients stick to treatment plans their providers prescribed, and answering everyday questions — "what do I do if I miss a dose?" — grounded in doctor-curated clinical knowledge, never improvised.

The hard constraint that shaped everything

Patient onboards Someone already under a clinician's care signs up.
EHR integration We verify against their electronic health record at onboarding.
Pre-diagnosed only Access is gated to patients with an existing diagnosis.
PHP assists Support the plan — never originate a diagnosis.
PHP never diagnoses. The EHR gate isn't a disclaimer — it's the architectural premise the whole product is built on.
Why this makes PHP a trust system, not a chatbot. By scoping out diagnosis and admitting only pre-diagnosed patients, every downstream decision — knowledge-graph grounding, doctor-in-the-loop evals, tiered action permissions — becomes a way to honor that boundary rather than a feature for its own sake.

What PHP was allowed to do

Support adherence

Keep patients on the treatment plans their providers already set.

Answer grounded questions

Everyday "what now?" questions, answered from curated clinical knowledge.

Act on the patient's behalf

Book appointments, fetch pharmacy coupons, order OTC wellness products.

The action capabilities had deliberate maturity limits — see Action tiers →.

The meta-problem — a mosaic of acquisitions

The company was a decade-long mosaic of acquisitions: six-plus business models, each holding a different slice of health value but living in its own silo. PHP's quiet thesis was to become the unifying conversational surface over all of them.

Healthlinehealth & wellness content
Healthgradesfind & rate doctors by zip
Optum Perkspharmacy coupons
Optum ProductsOTC e-commerce, delivered
PHP — one conversational surface over six products
Each acquisition became a tool the assistant could reach for — the mosaic is the substrate of PHP's action workflows, not just background.

My role

I owned the AI backend end-to-end — the conversational engine, the knowledge-graph RAG, the tool integrations, and the recommendation and action-completion workflows — leading a team of ~5 data scientists and running the doctor-panel evaluation program.