Orris Care — YC S2026 Application Evaluation

What I Read

Overall Rating: 5.5 / 10

The Company at a Glance

• Divyam Agrawal (CEO) — CA (All India Rank 31), Infosys finance associate, SF
• Takshit Shah (CPO) — MBBS doctor, rural medical officer, SF
• Akash Bhatiya (CTO) — Mechanical engineer turned software dev, Surat, Gujarat

Section-by-Section Breakdown

1. TEAM — Score: 6.5/10

• The founding team has a genuinely rare combination: a practising clinician (Takshit) who lived the problem, a CA who demonstrated hustling ability, and a self-taught dev. This is a real founder-market fit story.
• Childhood friends + 2 years working together — YC loves high-trust, known teams.
• All three are technical (self-taught coding) and all writing production code. That's a good signal.
• Takshit's MBBS + rural medical officer experience is directly relevant and compelling.
• Akash's prior work (Agrippon HIPAA-compliant billing, TestUI, real-time anonymous chat that spread by word-of-mouth) is strong — this is the most impressive builder on the team technically.

• Divyam's "most impressive thing other than this startup" is a red flag. He literally wrote: "to be honest the most impressive thing is actually my startup for which i am applying. I have been a little lazy for the times before that." YC's own guide says: "You probably shouldn't list the startup itself as your most impressive achievement." This is one of the most common rejection signals. It reads as self-awareness of a thin track record, and the honesty, while refreshing, doesn't compensate.
• Akash's "hack the system" story (PS5 arbitrage) is clever but minor — it's a fun anecdote, not extraordinary. Compare this to Divyam's sponsorship story, which is genuinely better structured.
• Equity split: Divyam 45%, Takshit 45%, Akash 10%. This is a significant problem. Akash is the CTO building production code. A 10% equity split to the technical architect who is also writing code looks like an afterthought hire, not a true cofounder. YC partners will notice this and wonder about team cohesion and whether Akash will stay through dilution.
• One founder (Akash) is still in Surat, India — while the application lists SF for the other two. Minor, but inconsistency in location details for a company claiming "San Francisco" base could raise questions.

2. IDEA — Score: 6/10

• The problem is real and the domain expertise is authentic. Takshit using GPT-wrappers that went viral (15k signups, 600 DAU) before building Orris is a powerful origin story — it demonstrates organic demand.
• The competitive analysis is honest and shows research (Reddit threads about OpenEvidence hallucinations, the RAG vs. fine-tuned model architectural debate).
• The 1,200-doctor interview claim is notable if true.

• The idea description is muddy. There's no single crisp sentence explaining what Orris does. The application jumps between "clinical decision support," "AI assistant for doctors," "can make PPTs," "write code," "generate posters," and "talk with medical books." YC wants matter-of-fact clarity. Right now this reads like a feature dump.
• The scope creep is a red flag. PPT generation, poster creation, code writing — these are not clinical decision support features. They suggest the team is building a general-purpose AI assistant, not a focused clinical tool. YC funds "vitamin to painkiller" products; listing PPT-making dilutes the painkiller story.
• No mention of the 15k signup viral moment in the main idea section. This is your single most compelling proof of demand and it's buried.
• The competitor section correctly identifies OpenEvidence but handwaves away the threat of ChatGPT/Claude. Those are real competitors in the "doctor's first instinct" battle — dismissing them without stronger evidence of why doctors switch to Orris specifically is weak.

3. PROGRESS & TRACTION — Score: 5/10

• Real revenue exists: $800 MRR, growing from $42 in November. That's roughly 19x revenue growth in 5 months — genuinely impressive.
• 1,500 DAU is real product usage, not just signups.
• 300 paying customers shows willingness to pay, which many early-stage apps lack.
• Delaware C-Corp is formed — structurally ready.

• $800 MRR with $2.50 ARPU is a serious concern. 300 paying customers at an average of $2.50 each means most people are buying $1-$3 credits, not recurring subscriptions. Only 20 users are on the $20 plan and 40 on the $10 plan. That's 60 recurring subscribers out of 300 "paying customers." This paints a picture of a largely transactional, low-commitment user base — not sticky SaaS.
• The growth trajectory is improving but the absolute numbers are very small. YC in 2025 expects "MVP in private beta" as table stakes. $800 MRR is below what competitive applicants are showing.
• "This excludes churn rate because within such a short time churn cannot be calculated reliably" — this line will concern YC. They will interpret it as: "we haven't measured churn." You should know your churn even if it's high. Saying you can't calculate it is a weaker answer than saying "monthly churn is X% but we've only had data for 3 months."
• The Demo Day goal is "$10,000 MRR." That's 12.5x current MRR in ~3 months. The path to get there isn't described. YC will ask: what's the growth lever? Paid marketing? Institutional partnerships? Sales motion?
• "Next stages are more dependent on growth in customer base and institutional tie-ups" — this is not a roadmap. YC wants to see a specific plan, not a hope.

4. IDEA MECHANICS / SCIENTIFIC BASIS — Score: 6/10

• The RAG + agentic while-loop + SQLite for scale approach is technically coherent and shows real engineering thought.
• Using Claude via AWS Bedrock is a solid, defensible choice.
• The multimodal capability mention (Gemini for image generation) shows product thinking.

• The technical section reads like a tech stack laundry list, not a defensible moat. "We use RAG" is table stakes in 2026 — everyone applying in healthcare AI is using RAG. What makes your RAG better than OpenEvidence's, or Cursor's medical mode, or Microsoft's Dragon Copilot?
• No mention of accuracy benchmarking. For a clinical decision support tool, the single most important technical question is: "Is it right?" There's no mention of hallucination rates, citation accuracy, or comparison against clinical gold standards.
• The regulatory answer — "No approvals required as of now since our application is a suggestive layer" — is accurate legally but sounds dismissive. YC partners know healthcare is a regulatory minefield. A better answer acknowledges the FDA's AI/ML guidance and explains why the "clinical decision support" exemption applies to your specific use case.

5. MARKET — Score: 7/10

• "1 million physicians in the US, expanding to nurses, PAs, and other clinicians globally" is a real TAM.
• Healthcare AI is explicitly on YC's Requests for Startups for 2025.

• The market sizing is stated without numbers. What's the TAM in dollars? YC wants to see "$X billion addressable market" with a bottom-up logic.
• Individual clinician subscriptions at $10–20/month is a hard GTM. The real money in healthcare AI is in B2B (hospital systems, payers, pharma). There's no mention of an enterprise path.

6. APPLICATION WRITING QUALITY — Score: 4.5/10

• Grammar and polish issues throughout. Inconsistent capitalization ("$" followed by spaces, "hitps://" typos in URLs, lowercase proper nouns). YC partners read on phones and these friction points add up.
• Marketing language appears. Phrases like "highly reachable milestone" and "this would be profitable and scalable" are the exact buzzwords YC's guide says to eliminate.
• The "lazy" admission in Divyam's section is self-sabotaging. Honesty is good; self-deprecation about your own work ethic on a YC application is not.
• Answers frequently start with lowercase letters and lack proper sentence structure.
• The Agrippon/TestUI work by Akash is excellent but is buried in a long background section. It should be surfaced more prominently.

Top 7 Things to Fix Before the Next Batch

What's Genuinely Strong

1. Founder-market fit is real. A practising MBBS doctor as CPO who personally lived the problem and built early GPT tools that went viral — this is exactly what YC looks for. This is your single best asset.
2. Early traction exists. $800 MRR growing from near-zero in 5 months, with real paying users, is better than most pre-revenue applications.
3. The team is high-trust and committed. 2+ years together, all full-time, childhood friends — this reduces the cofounder-risk signal.
4. Healthcare AI is on YC's radar. Multiple healthcare-IT companies (Penciled, Locata, Clarion, Copper Health) have been funded in recent batches. You're in a hot category.
5. The viral origin story (15k signups, 600 DAU on a basic GPT wrapper) is a powerful proof of demand — it just needs to be front-and-center in the application.

Final Verdict