Who I Am
19-year-old CS student at Gujarat Technological University, graduating 2027. That's the simple version. The fuller version: CTO at Knight Medicare (healthcare AI startup), Local Director at Open Paws India (animal welfare org), founder of Open Permit (civic tech) and Adventurers Guild (edtech), Google Gemini Campus Ambassador, and active contributor to open source repos with 1k–3.6k stars.
I'm not writing this to brag. I'm writing it because the "how do you manage all this" question gets asked constantly, and the honest answer is more nuanced than the LinkedIn-friendly version. There's a real system underneath — and also real failures, real burnout, and real chaos. Both are true simultaneously.
This post is the unedited version.
The Setup
Four primary roles, running in parallel, with different accountability structures, different stakeholders, and very different work types. Each demands different cognitive modes.
CTO — Knight Medicare
Healthcare AI startup. Technical architecture, roadmap, and engineering execution. Real product, real patients downstream, real stakes. The hardest role to context-switch out of because the systems are genuinely complex and bugs have consequences.
Local Director — Open Paws India
Organizational leadership for Open Paws India operations. Community management, event coordination, team building, stakeholder relationships with global Open Paws leadership. More people management than code.
Founder — Open Permit + Adventurers Guild
Two products: Open Permit (civic tech, permit objection AI) and Adventurers Guild (CS education RPG). Both early-stage, both need product decisions, both need code. The founder role means being the last person responsible for everything in the build.
Student — GTU, Batch 2027
CS degree. Classes, exams, labs, projects. Cannot be fully deprioritized — the degree matters, the credits matter, the finals happen whether or not a product launch is also happening that week. The role that least tolerates flexibility.
How I Actually Manage This
The honest answer: Claude Code is how. 6-7 sessions per day. 4-5x force multiplier on output. Without it, I'd run 2-3 projects max — not 11. This isn't an exaggeration; it's the actual arithmetic of how much gets done per hour with and without AI-assisted development.
But tooling alone isn't the answer. The structure underneath matters too. A typical day runs in four modes — each protecting a different type of work from context-switching contamination.
Morning: Code Reviews + Pipeline Checks
Start with the things that can block others. Review PRs, check deployment health, resolve any overnight issues in production systems. Low cognitive load, high unblocking value. The goal is clearing the queue before doing deep work.
Afternoon: Architecture + New Features
The best cognitive hours go to the hardest technical work. New feature design, architecture decisions, complex debugging sessions. Claude Code sessions during this window are the most productive — the quality of the session depends on the quality of the thinking going in.
Evening: OSS Contributions + College Work
OSS contributions are evenings and weekends — the work is self-directed, the stakes are lower if interrupted, and the reviewing maintainer is often in a different timezone anyway. College assignments here too. Not ideal, but it works.
Async: Pitch Decks + Funder Outreach
Everything that doesn't require synchronous focus — funder emails, pitch deck iterations, Brevo campaign management, documentation. Batched and done in async blocks rather than interrupting deep work time. Mostly Claude Code sessions driving drafts, me steering and approving.
Wins Worth Noting
Not humble-bragging — recording these because the system needs to produce real outputs or it's not a system, it's just busyness. These are the markers that confirm the model works.
Things That Went Wrong
The system has failure modes. These aren't hypothetical — they happened. Putting them here because the "building in public" framing is meaningless without the failures.
Running many projects doesn't mean running them perfectly. Here are four times the wheels came off:
Supabase India Block — 3 Days Down
Supabase blocked Indian IP addresses with no warning. Production down for 3 days across multiple services while migrating infrastructure. Every project that relied on Supabase was affected simultaneously. The fix (Yotta migration) was the right long-term call, but those 3 days were brutal. Lesson: single-vendor dependency on cloud infrastructure is a risk you can't ignore.
Missed Hackathon Deadline
Timezone confusion on a submission deadline. Built something real, shipped nothing. The kind of mistake that's preventable with basic calendar hygiene, which I didn't have at the time. Now every hackathon deadline is in the calendar 3 days early with a reminder. Simple fix, should have been in place already.
Underestimated AARC Pitch Prep
The AARC Pre-Accelerator pitch required more preparation than I estimated. Showed up with good product, undersold it with insufficient pitch polish. Investor pitch is a specific skill that doesn't transfer directly from "can explain the product to a developer." Spent two weeks after that doing nothing but pitch iteration — wouldn't have needed to if I'd planned ahead.
Burnout (College Finals + Product Deadline)
GTU finals and a Knight Medicare deployment landed in the same week. Not a scheduling accident — both had fixed dates that happened to overlap. The week was survivable; the week after was not. Output dropped to near zero for 5 days while recovery happened. The system doesn't have enough buffer for simultaneous hard constraints. Still working on this.
What's Ahead
Five concrete milestones in the next 12 months. Not aspirations — actual targets with defined success states.
Open Paws India Pvt Ltd Registration
Formalizing the India entity for legal operations, banking, and grant eligibility. Currently operating as an informal local chapter; registration unlocks funding sources and government partnerships that require a registered entity.
Knight Medicare Mental Health Org Pilot
First live pilot with a mental health organization using Knight Medicare's AI triage and documentation tools. Moving from internal testing to real clinicians and real patients. The pilot that proves the product works in production.
Open Permit Fundraising Round
SFF grant application and first angel round for Open Permit. The product is post-AARC, has 8+ country coverage, and has validated demand. The round funds the infrastructure expansion and dedicated legal framework research capacity.
Adventurers Guild 1000 Users
From 500 to 1000 active users. Growth is primarily referral-driven right now; scaling to 1000 requires a more systematic acquisition channel — probably university partnerships and GTU ambassador network leverage.
LEWS Methodology Research Paper
Formalizing the LEWS scoring methodology as a publishable research paper. Dataset exists from the EA Animal Welfare Hackathon. Co-authors (Tim, Thomas, Rapha) are intact. The paper documents what we built so others can replicate it.
Key Lesson
The secret is not being exceptional. It's having systems. Systems beat talent when talent doesn't have systems. Consistent daily execution inside a clear structure compounds faster than occasional bursts of heroic effort. The 6-7 Claude Code sessions, the morning/afternoon/evening structure, the async batching — none of it is clever. It's just doing the same things every day until the outputs accumulate.