Hey, I'm Sarthak!
I build AI that grows companies.Founder's Office at FloBiz (myBillBook), ex-Amazon. My AI agents find revenue, answer customers, and run the content engine behind 150M views in a single month.
Reach
150M views
in one month, on a ₹5L budget
ARR
52% ARR growth
YoY, up from 23%. Monthly plans were my call.
Savings
₹20L saved / month
with AI products and SOPs I built
AI team
Me + 50 AI agents
my analysis, QA and reporting team
What I own, what I've built, and who I am off the clock
From Amazon to the Founder's Office
At Amazon I owned network-loss analytics for India and four international markets: 50+ loss buckets, tracing where money leaked out of the supply chain. I went from a 6-person analytics team to running a program of 60, and drove ~₹2M a week in direct savings.
At FloBiz I sit in the Founder's Office and own the numbers the founders decide on. I built the revenue attribution engine that matches the sales team's books to the rupee, automated the investor ARR report, and pitched the move to monthly plans. ARR growth went from 23% to 52% year on year. One read found that 37% of users up for renewal bring 82% of renewal revenue, and the rule held on data it had never seen.
Then I got tired of doing it by hand.
So I built AI to do it, and I check every number it gives me.
AI does the repeat work. I check every number.
Most of my work runs through Claude Code. I run 50 agents, 25 skills I wrote myself, and 13 hooks. The hooks block a secret in code or a sensitive term in a commit, and make me prove a claim before it goes to another team. A 500-file memory carries every decision from one session to the next.
Flo Analytics: my plain-English-to-SQL engine and the MCP server behind it, with 23 tools over company data. It powers our customer-facing assistant: 127 myBillBook users asked it 477 questions in 30 days. I run a 1,062-question regression set before I ship a change.
Support: an AI auditor that grades support chats and scores the bot and each human agent separately. The LLM gives verdicts and quotes the evidence; Python does every sum. Next to it, a WhatsApp bot that answers only from our help docs, projected at about ₹55 per 1,000 conversations.
Cost: the AI products and SOPs I built save about ₹20L every month. I also mapped the company's vendor spend to what in-house AI can replace next. Separately, I tested a voice-AI model for collection calls that costs about 5× less per minute.
I've built auditors, SEO tools, WhatsApp bots (including group expense tracking), and full-stack AI systems.
The first job I automated was mine.
Then I went looking for the next one.
Jalpara: where I think best
I call myself Jalpara: a man who belongs in water. Put me near a river, a backwater or a waterfall and I'll find my way in. I've walked hidden trails to reach water and swum in freezing cold.
Near water, my head goes quiet.
That's where I do my best thinking.
Sport. Trade. Languages.
I play everything: football, cricket, badminton, golf, squash, TT. And I watch everything: the Olympics, World Cups, league nights. I rarely miss the moments that matter. Huge Virat Kohli fan.
I trade 6E futures with a scanner I built. It reads volume and sentiment from five sources across several timeframes, and my Apple Watch logs how my body reacts to each trade.
I speak Hindi, English, Sanskrit and French, and I'm learning Spanish and Punjabi.
150M views on ₹5L
I head the content team at FloBiz. In September 2026 we did 150M views on ₹5L spend: about ₹3.3 per 1,000 views. We built it to run every day, not to bet on one viral reel.
Four engines feed it: creator reels that the team commissions, in-house channels, reposts, and AI-made content. My Repost Portal turns a reel into a branded, credited repost with about 30 seconds of human time. A view tracker reads every reel every day.
I measure it like revenue. Branded Google searches grew about 22% above the trend line after the programme started.
I don't over-plan. I build, test, refine, and ship.
And when I say ship — I ship fast, and I ship often.
Where I've worked, and what I shipped there.
FloBiz (myBillBook)
Oct 2024 – Present
Founder's Office · Strategy & Special Projects Lead
Revenue. Built the attribution engine that assigns every payment to a team every 15 minutes and matches the sales team's books to the rupee. It found an in-app product line the sales revenue tracker had been leaving out. Automated the investor ARR report with blocking reconciliation checks, which caught silent errors the manual process had let through.
Decisions. Pitched the move to monthly plans and built the data case for it; ARR growth rose from 23% to 52% year on year (Sep 2025 → Sep 2026). Pricing and renewal reads the founders act on. 37% of users up for renewal bring 82% of renewal revenue, and the rule held on unseen data. Showed leadership that a miss against the annual plan was timing from a deliberate shift to monthly plans, not lost demand.
AI in the product. A plain-English-to-SQL engine and an MCP server with 23 tools over company data. It powers our customer-facing assistant: 127 users asked it 477 questions in 30 days.
Cost and support. The AI products and SOPs I built save about ₹20L a month. They include AI that grades support chats, scoring the bot and each agent separately, and a WhatsApp bot that answers only from our help docs, now in testing at a projected ~₹55 per 1,000 conversations.
Content. Head the content team: creator reels, in-house channels, reposts and AI. 150M views in September 2026 on ₹5L.
FloBiz (myBillBook)
Oct 2024 – Present
Founder's Office · Strategy & Special Projects Lead
Revenue. Built the attribution engine that assigns every payment to a team every 15 minutes and matches the sales team's books to the rupee. It found an in-app product line the sales revenue tracker had been leaving out. Automated the investor ARR report with blocking reconciliation checks, which caught silent errors the manual process had let through.
Decisions. Pitched the move to monthly plans and built the data case for it; ARR growth rose from 23% to 52% year on year (Sep 2025 → Sep 2026). Pricing and renewal reads the founders act on. 37% of users up for renewal bring 82% of renewal revenue, and the rule held on unseen data. Showed leadership that a miss against the annual plan was timing from a deliberate shift to monthly plans, not lost demand.
AI in the product. A plain-English-to-SQL engine and an MCP server with 23 tools over company data. It powers our customer-facing assistant: 127 users asked it 477 questions in 30 days.
Cost and support. The AI products and SOPs I built save about ₹20L a month. They include AI that grades support chats, scoring the bot and each agent separately, and a WhatsApp bot that answers only from our help docs, now in testing at a projected ~₹55 per 1,000 conversations.
Content. Head the content team: creator reels, in-house channels, reposts and AI. 150M views in September 2026 on ₹5L.
Amazon
Jan 2022 – Oct 2024
Program Manager / Business Analyst
Owned network-loss analytics for India and four international markets (Japan, Saudi Arabia, UAE, parts of the US). Led a 6-person analytics team, then moved to program management and ran a team of 60. Drove ~₹2M a week in direct savings from loss reduction.
Amazon
Jan 2022 – Oct 2024
Program Manager / Business Analyst
Owned network-loss analytics for India and four international markets (Japan, Saudi Arabia, UAE, parts of the US). Led a 6-person analytics team, then moved to program management and ran a team of 60. Drove ~₹2M a week in direct savings from loss reduction.
UpSchool
2018 – 2020
Founder
Founded an EdTech platform in college that connected students, parents and schools. Built and ran it for 2+ years alongside my degree at DTU.
UpSchool
2018 – 2020
Founder
Founded an EdTech platform in college that connected students, parents and schools. Built and ran it for 2+ years alongside my degree at DTU.
DTU
2018 – 2022
BTech, Math & Computer Science
General Secretary of E-Cell. Interned at Axxela trading US 10Y T-Note futures. Volunteered with United Nations and India Against Covid.
DTU
2018 – 2022
BTech, Math & Computer Science
General Secretary of E-Cell. Interned at Axxela trading US 10Y T-Note futures. Volunteered with United Nations and India Against Covid.
Things I've built that people actually use.
Flo Analytics
Most people can't write SQL, so every data question waits for an analyst. I built an engine that takes the question in plain English: question in, SQL out, answer and chart back. It ships as an SDK and as an MCP server with 23 tools over the company's warehouse, product analytics and CRM.
It powers our customer-facing assistant: 127 myBillBook users asked it 477 questions in 30 days. I run a 1,062-question regression set before I ship a change, so a fix for one question never breaks another.
Support Automation
An AI auditor that grades support chats and scores the bot and each human agent separately. The LLM only gives verdicts and quotes the evidence; Python does every sum, so a warm, polite chat can't hide a failed check. It traced the most common auto-fail to a handoff template, not to the agents.
Next to it, a WhatsApp support bot that answers only from our help docs, projected at about ₹55 per 1,000 conversations. Building it surfaced 65 contradictions in the help docs themselves.
Content Engine
How we got 150M views in September 2026 on ₹5L: creator reels, in-house channels, reposts and AI content, run by the team I head.
The Repost Portal turns any reel into a branded, credited repost with about 30 seconds of human time: 7 layouts, 19 add-ons, 550+ videos rendered so far. A view tracker reads Instagram, YouTube and Facebook every day, and a model ties views to branded Google searches.
Archana
Booking a pooja online needs trust first. I built a trust-first pooja booking platform with photo proof of every pooja, daily Panchang, Kundli birth charts and auto-refund via Razorpay.
Live at myarchana.in and taking real payments, with an A/B price test running now. Zero paid APIs: Panchang and Kundli are computed on my own server. 350+ commits.
Site Auditor
Any URL to an AI UX audit in 60 seconds. Lighthouse scores plus 4 user personas across 8 UX pillars.
Social Tracker
Multi-tenant SaaS that reads reel views on Instagram, YouTube and Facebook every day, with monthly cohorts.
myVoiceBooksAI
Voice-first analytics for small shop owners. Ask in Hindi or English, get the answer and a chart.
ToyHype
Cash-on-delivery-first e-commerce for India. Storefront, coupons, order tracking, Razorpay and Shiprocket.
Task Dashboard
Multi-team task board: drag-and-drop columns, Google Calendar sync, GitHub OAuth, XP and streaks.
VidText
Transcribe and summarize YouTube, Instagram reels and Spotify podcasts. Chrome extension plus Kindle delivery.
Morning Coffee
One morning briefing from my calendar, Gmail, Slack, git activity and trading levels.
Pulse Check
Scans every project I run: git activity, Slack DMs, stalled work. Returns one ranked punch list.
Tools anyone can use today
Products I run in public. Try them. That's the best way to see how I build.
Archana
Live productBook a pooja online with photo proof and auto-refunds. Daily Panchang and Kundli charts, free for anyone. Live and taking payments.
Repost Portal
SaaSPaste a reel, pick a layout, get a branded and credited repost back. 7 layouts, 19 add-ons, about 30 seconds of human time per video.
Site Auditor
Free toolEnter any URL and get a UX audit in 60 seconds: Lighthouse scores, 4 user personas, 8 UX pillars, and code-level fixes.
Social Tracker
SaaSTrack reel views across Instagram, YouTube and Facebook without a spreadsheet. Daily refresh, monthly cohorts, multi-brand.
Stay in the loop
I write about building AI products, shipping fast, and lessons from the trenches. No spam, unsubscribe anytime.
Get in touch
Founder's Office and AI product roles, or a contract to ship one AI system end to end.
If a number on this page made you curious, ask me how.
Based in Delhi, India.
Open to remote.
© 2026 Sarthak Goel