Associate Product Manager, Founder's Office at Visa2Fly. The $10M ARR B2B SaaS visa platform with 700+ paying agency tenants, backed by Flipkart Ventures, M Venture Partners, and FinSight Ventures. The same product layer also powers B2B2C consumer-platform integrations and our B2C direct journey across 75+ countries.
About
I joined Visa2Fly in June 2022 as employee #11. Started in ops, opened four South India cities as EIR, then built the analytics layer the company runs on, and now own product. Four years. Three roles. Every system, dashboard, and product surface I'm responsible for today, I built or rebuilt myself.
Today I'm Associate Product Manager in the Founder's Office, reporting to the co-founder who serves as CTPO (Chief Technology & Product Officer). I own the product roadmap end-to-end and ship with engineering, design, and AI; partner with ops, finance, and legal as projects demand. Visa2Fly is a B2B SaaS at its core. 700+ paying agency tenants now drive 75% of the $10M ARR. The same product layer powers B2B2C consumer-platform integrations and our B2C direct journey.
I read the dashboards before I write the spec. Operator first, builder always.
Experience
Employee #11. Hired and led a 4-person ops team. Stood up Visa2Fly's first SLA framework, vendor onboarding, and process documentation. Led SEA region visa operations with channel partners. EIR · South India expansion (Oct 2022): opened Chennai, Bangalore, Hyderabad, and Mumbai; 35% CAGR post-launch.
Featured product
The visa product runs end-to-end on AI. 4x throughput on the same team, 96% faster cycle time, 68% lower cost-per-visa. Scroll through what's inside.
Document intake runs through a tuned OCR layer that pulls passport, photo, and form fields from scans and phone photos, even crumpled ones. Fields normalize into a schema the rest of the product trusts.
A Vision-Language Model verifies every OCR pull: matches passport photo to form, flags mismatched names, spots blurry pages, reads handwritten edits. Anything it can't confidently validate is routed to a human before we touch a government portal.
Automated bots file the validated application into each country's portal, handling session state, captchas, and varying form shapes. What took agents 20 minutes of copy-paste now finishes in under 3.
Payment orchestration triggers at the right handoff (government fees, service fees, refunds on rejection) with reconciliation back to finance. No stuck applications waiting on a payment ticket.
Once a visa is granted, the product downloads the approval, packages it with travel docs, notifies the customer, and closes the loop in CRM. SLAs track every stage, so ops sees red the moment something slips.
Selected work
Five chapters at Visa2Fly: open new markets, measure what works, solve the consumer journey, fix the unit economics, turn it into a B2B SaaS the industry runs on. Each case lays out the industry problem, the goal, the competitive context, what I learned from customers, what I decided (and what I said no to), what shipped, and the trade-off I'd defend.
Problem: The visa-reseller industry runs on WhatsApp. Big agencies had hacky custom tools; small ones had Excel. Visa2Fly's V1 dashboard existed but was utility-grade: agencies closed visas through it, but didn't run their business on it. Goal: Turn the dashboard from a filing tool into the operating system every agency runs its visa business on. Market & Competition: No competitor offered margin tracking, commission flows, or add-on services natively in a visa platform. Agencies were patching this together in spreadsheets.
Discovery: From 30+ partner discovery calls and on-floor observation: bulk groups of 5 to 500 travelers were the real workflow, not individual filings. The biggest missed lever was revenue. Agents had no way to attach insurance, and margin discipline was a manual spreadsheet exercise. Decision: Reframe the dashboard. Ship revenue surfaces (markup, commission, add-on services) and speed surfaces (bulk + auto-mapping + live tracking) in the same release. Shipped: bulk upload of 100+ travelers with auto-mapping under 15 seconds (fastest in the travel industry); live application tracking with real-time government-side stage updates; dynamic country-specific checklists; automated cover letter and itinerary generation; markup + commission management; add-on services at point of sale (insurance); wallet with low-balance alerts; partner margin view; 24×7 chatbot across 75+ countries. Trade-off: Invested the engineering quarter in the B2B partner surface over consumer brand build-out. The 25% → 75% B2B share of business validated the bet.
Next timeA self-serve activation flow so the SMB long-tail onboards without a sales call. We hand-held the first 500 partners.
View partner dashboardProblem: Filing eVisas was the cost-per-visa bottleneck for every visa company. Industry average sat around $1+ per filed visa; at scale that was the unit-economics ceiling. Customer-facing time-to-apply sat at 15 hours. Goal: Cut cost-per-eVisa to a level that supports growth across all target countries at zero ops headcount additions. Market & Competition: Most players ran 1 to 2 OCR layers without multi-stage VLM validation. Industry-average human-review rate was 15 to 20%.
Discovery: Time-and-motion audit showed 70% of filing time was repetitive entry the human added zero value to. Errors clustered at document interpretation (passport variations, photo specs, partial scans). The 30+ high-volume countries we automate drive 85% of business. Decision: Buy what's commoditized (OCR + VLM model layer), build what's our edge (validation, confidence thresholds, multi-channel orchestration). Automate the 30+ countries that drive 85% of business deeply, instead of spreading thin across the long tail. Owned the buy-vs-build call with the founders. Shipped: third-party OCR + VLM APIs wrapped in a Visa2Fly-built validation, escalation, and orchestration layer. The models are vendor; the orchestration around them is ours. Multi-channel delivery. PRD with the co-founder (CTPO); weekly working sessions with the AI engineer on confidence thresholds; ops aligned on the 2% human-review SOP. Cost per eVisa $1.00 → $0.25, customer-facing TAT 15 hours → 1 hour, 4× throughput on the same ops team, 99.3% accepted on first government submission, human review under 2% of cases. Trade-off: Concentrated the build on the 30+ high-volume countries, leaving the long tail on manual ops. The 80/20 unit-economics math justified the focus.
Next timeStart with a 5-country pilot to lock confidence thresholds before parallel rollout. We scaled in parallel and had to retune per-country mid-flight.
View AI architectureProblem: Sticker visas were the most underserved corner of the visa market. Travelers piecemealed embassy slot booking, document chaos, and 8-step paperwork on their own. Document validation TAT sat at 48 hours; escalation miss rate ran at 2.5%. Goal: Build the first unified, fully automated Sticker Visa journey in the industry. Market & Competition: Four leading visa-tech players reviewed. None offered prep checklists, live status, country-purpose specific flows, or end-to-end logistics.
Discovery: 25+ consumer interviews + 12 partner-agency reviews. Drop-off concentrated between document upload and payment. The checklist needed to be country AND purpose specific (tourism, business, family), not generic. Agents wanted logistics handled: pickup and drop of physical documents. Decision: Redesign Sticker Visa as a fully automated smart journey. B2B partner-led launch first (the agency does the heavy lift); B2C as halo. Build dynamic checklists (country × purpose). Integrate logistics APIs so the system drives pickup and drop end-to-end. Shipped: dynamic country × purpose checklists; automated document validation 48 hours → 2 hours; live consumer-facing stage tracking; logistics API integrations for system-driven auto pickup and drop; escalation miss rate 2.5% → 1%; journey compressed from 8 broken manual steps to 5 fully automated smart steps; self-service post-purchase. Live across 45+ countries (UAE, KSA, SG, ID, Schengen). Trade-off: Launched all 45+ countries simultaneously instead of one-by-one. Per-country edge cases absorbed by the post-purchase dashboard without breaking the funnel.
Next timeAdd a save-and-resume layer for users who pause mid-flow. About 15% restart from scratch today; that's the largest remaining drop.
View booking flowProblem: Funnel decisions ran on gut. Three tools held three views of leads. Marketing couldn't see what was converting; ops didn't see SLA breaches until customers complained. Cost of bad calls compounding weekly. Goal: Move the org from gut-driven to test-driven funnel decisions. Market & Competition: Internal product, no external benchmark. Closest analogs (Optimizely, GrowthBook) were overkill for a 50-person team running a single-product funnel.
Discovery: Routing audit showed round-robin was sending high-value leads to cold reps. The 50-person team rerun the same lookups every morning across tools. Score-ranked routing was the obvious lift, if we had infrastructure to test it. Decision: Ship the experimentation framework before the dashboards. Tooling-first means every future decision is testable instead of retrofitted. Shipped: org-wide A/B framework (score-ranked routing, 2,000 leads per arm); unified Sales + Ops product with 5 SLAs and breach alerts. Defined the data model with engineering, ran weekly metric reviews with founders and sales leadership, trained ops + sales on dashboard self-service. +22% conversion lift, statistical significance, 27% faster customer connect, 7% ops efficiency. Trade-off: Quicker wins were available if we'd built dashboards first. Worth it: every funnel decision since runs through the framework.
Next timeWrite experiment templates earlier. The first three months teams ran ad-hoc tests we had to re-run later with proper controls.
Read moreProblem: Visa2Fly worked in Delhi NCR but had no playbook for replicating outside. The visa industry hadn't solved city expansion either: every launch in travel-tech tends to fragment ops, customer experience, and quality. Goal: Build the first repeatable city-launch playbook in visa-tech. Market & Competition: No template existed in the industry. Closest analog: travel-aggregator city launches (MakeMyTrip, Cleartrip), which fragmented quality across cities.
Discovery: Two weeks of on-ground visits to Bangalore and Chennai. Vendor networks varied wildly. Some cities had reliable couriers, others ran on local agents on motorbikes. The playbook had to abstract that variability or break. Decision: Ship process docs over building city-specific tooling. Speed-to-launch beat scale-now. 11-person company, runway not a platform team. Shipped: city-launch playbook (vendor onboarding, SLA framework, field-agent coordination); 4-person ops team I hired and led; SEA region operations with channel partners. Live in Chennai, Bangalore, Hyderabad, Mumbai. Foundation that scaled to 700+ partners three years later. Trade-off: Tooling would have scaled better long-term. We chose 2 to 3 months of speed over months of platform build.
Next timeThe playbook held up. I'd version-control it in Notion so the next city launch could fork it instantly.
Read more3 wireframes · booking flow, AI architecture, B2B partner dashboard
Skills
End-to-end product work: PRDs, experimentation design, unit-economic gates. I ship product, not slides.
Maharaja Surajmal Institute, GGSIPU, Delhi
Let's talk
Hiring an APM, PM, or Founder's Office product hire? I'd love a 20-minute chat.
Contact
Role, collaboration, or just a coffee chat about AI products. I'm always up for a good conversation.
What I'm looking for: APM / PM roles at growth-stage B2B SaaS in India, UAE, Singapore, or remote.