TL;DR
- MediSwift was an AI assistant that collected patient history, transcribed doctor–patient conversations, and generated FHIR‑compliant documentation.
- In 2 months we shipped an MVP that ran inside two high‑footfall clinics in India.
- We shut it down nine months later because the operational distance (I was in Germany while our target users were in India) made scaling impossible.
- Handing the codebase to teams on the ground felt bittersweet—but I left with sharper instincts and a long list of lessons for my next venture.
The Spark
I live with G6PD deficiency, psoriasis, and the occasional kidney stone. Each new doctor visit forced me to retell my medical story from scratch—even though the data already existed somewhere in a file or portal. In March 2024, that frustration finally tipped into action. Why isn’t this seamless? MediSwift was born from that single, patient‑side question.
Building at Warp Speed
-
Place & timing: I incorporated MediSwift while living in Stuttgart, Germany, but aimed squarely at Indian clinics—my first strategic misstep.
-
The team:
-
Rushab Ghoghari – CTO, product development (AI & infra)
-
Manshi Ghoghari – COO, partnerships & investor outreach
-
Me (Pranav) – CEO/engineer, product vision & go‑to‑market
-
Tech stack: Web‑based app → Whisper for speech‑to‑text → custom speaker‑diarisation → fine‑tuned GPT‑4o that outputs FHIR‑compliant JSON.
-
MVP in 60 days: Two general‑physician clinics agreed to pilot because documentation was their #1 pain. Seeing elderly patients chat with our intake bot while the doctor’s notes appeared auto‑magically was surreal.
Small Wins That Felt Huge
- Watching a 70‑year‑old patient finish the chatbot intake without help.
- A doctor telling us, “You just gave me an extra hour to breathe.”
- Acceptance into Startupbootcamp (Techstars & YC interviews were close but no cigar).
The Wall We Hit
“Build where you live, or live where you build.”
Operating from Germany while selling into India was like driving with the handbrake on:
- Time zones & trust – Doctors wouldn’t test new software outside clinic hours, and my evenings vanished into support calls.
- Regulatory maze – Compliance reviews required in‑person hospital visits and local legal counsel.
- Personal bandwidth – A new marriage and lingering student loans meant relocating to India wasn’t realistic.
Shutting Down—Gracefully
By late 2024 it was clear we’d need an on‑site team and fresh capital to scale. Rather than raise money we couldn’t deploy well, we chose to wind things down.
Steps we took:
- Told the pilots first – honest email + Zoom call; offered full data export.
- Team huddle – acknowledged the runway reality; helped each member plan next moves.
- Hand‑off – shared the codebase and roadmap with two India‑based founders tackling the same space.
Switching off servers, canceling API keys, and archiving thousands of lines of code hurt. But relief followed quickly: the lesson was banked, the guilt was gone.
Five Things I’d Do Differently
- Live near your first users—especially in healthcare, where trust is local.
- Choose co‑founders for complementary time‑zones and networks, not just skill sets.
- Validate sales cycles early. A demo is fun; a signed LOI at realistic prices is gold.
- Keep infra cheap until usage forces you to scale. Whisper + GPT‑4o is powerful but pricey if idle.
- Build for one clear persona first. General physicians ≠ specialists ≠ telehealth.
What’s Next
I’m back in Germany, working on AI product strategy while sketching new ideas. If you’re tackling clinical documentation—or just want to trade war stories—reach out.
Closing Thought
Startups end, but learning compounds. MediSwift taught me more in nine months than three years of comfortable employment ever could. I’m grateful for every late‑night deploy, every doctors’ lounge demo, and even the shutdown forms.
Onward, Pranav Ghoghari pranavghoghari.com