rohit gawli
data guy in nyc, leading analytics engineering in adtech and teaching ai agents to speak proprietary data. before that: two startups in mumbai, one exit.
select * from numbers_i_stand_behind;
| solo_founder_revenue-- trillium, zero employees | $0.5m | |
| livelihoods_created-- lokal kitchen, home chefs earning | 500+ | |
| ngos_supported-- furtribe, on a $0 budget | 40+ | |
| founder_exits-- trillium, 18 months in | 1 |
rabbit_holes
public-data research on real-world questions, with the receipts included.
company_brain
how to build a company brain: one Markdown file per metric, an index in every folder, and an agent that opens only the files the question needs, so "what was our trial-to-paid conversion last month?" gets the same answer every time.
→
startup_graveyard
a census of all 3,566 YC companies from the 2021–2026 batches — survival curves, 300 shutdowns, and the receipts for how they died.
→
the_hiring_signal
what 151 nyc ai startups actually hire for — 522 public roles, classified & mapped.
→
all rabbit holes →
projects
📊
analytics_engineering
canonical data layer, dbt + airflow, org-wide standards.
🤖
ai_agents
agents fluent in proprietary data: ai brings the proof, humans make the call.
🐾
furtribe
open-source animal welfare platform, $0 budget.
→
🍛
lokal_kitchen
zero-fee home-chef marketplace, mumbai.
→
🧴
trillium
d2c personal care, $0.5m revenue as solo founder, exited in 18 months.
writing
when code gets cheap, the question gets expensive
→
the software moat moved up the stack
an agent connected to a warehouse can answer a question about “active customers” without knowing what an active customer is.
→
the agent needs a library before it needs more power
a question arrives: “how many active customers did we have last quarter?”
→
what most analysts get wrong about their own career
being right is not a career strategy.
→
more on medium →