Selected work.
Six engagements that say what we do better than a deck would. Two are public; the rest are anonymised at our clients' request.
◉ MVP2025MVPMarketingTravel · Website + CRM · 2025Tripceler
Role · Product, engineering, design, growthThe problem.An independent travel company was bleeding leads in a stack of spreadsheets, WhatsApp threads, and a brochure-ware website. Every enquiry was a manual relay race, and the slowest leg lost the booking.
What we did.We rebuilt Tripceler.com from scratch as a destination-first travel website with a CRM brain. Every enquiry — homepage form, package page, chat — drops directly into a structured pipeline, gets auto-assigned, and starts a nurture sequence before the lead closes the tab. Turnaround on first response collapsed from hours to minutes.
Lead → first reply: <5 minSingle source of truthCustom CRM, zero plugins
◉ AI2025AIMVPFinTech AI · GST reconciliation · 2025Project Ledger
Role · AI architecture, product, full-stack engineeringThe problem.A senior CA firm in India was losing hundreds of partner-hours every month to GST reconciliation. The rules shift quarterly, the data is messy, and one mismatched invoice can unravel a return. The default tools were either too literal or too generic.
What we did.We built a SaaS reconciliation engine wired to the latest GST rule-set. Invoices, GSTR-2A/2B, and ledgers are ingested, normalised, and matched by an LLM-assisted rules layer with confidence scoring and human-in-the-loop review on the edge cases. Work that used to take 50–60 hours per cycle now closes in 2–3.
~20× faster close cycleRule-set auto-updatedConfidence-graded matches※ Client name under NDA. Engagement ongoing.
◉ Marketing2025MarketingVideoD2C · Media + Growth · 2025Northwind Co.
Role · Creative direction, paid media, lifecycleThe problem.A young D2C brand had a great product, a CAC that wouldn't sit still, and a creative pipeline that ran dry every 10 days. They needed one team to own the work end-to-end.
What we did.We ran a single pod across creative, paid, and lifecycle — a quarterly cadence of brand films plus a rolling library of platform-native cuts. First-touch profitability stabilised inside three weeks; CAC trended down month-on-month from week four.
Steady-state CAC ↓One pod, full funnelQuarterly creative cadenceReference on request
◉ AI2025AIMVPLogistics · Internal AI copilot · 2025Stride Ops
Role · Multi-agent design, RAG, evalsThe problem.A mid-market logistics operator wanted to give their dispatch team an AI copilot, but every off-the-shelf assistant fell over on the actual playbook — naming conventions, regional carve-outs, exception flows.
What we did.A planner/tools/critic agent stack with retrieval over six years of routing data and dispatch SOPs. Eval-gated deploys, human-in-the-loop on anything touching a customer commit. Pilot landed in eight weeks; expanded twice since.
−43% routing time+11% on-time rateEval-gated deploys※ Anonymised at client request.Reference on request
◉ Video2024VideoCreator · Brand film · 2024Verge Studio
Role · Direction, production, postThe problem.A creator-led brand needed a launch film that worked as a 180-second hero and as 20+ feed-native cuts, without three vendors and four months.
What we did.Single brief, single team, two cities. Shot in Bengaluru and Lisbon over a week; delivered the hero film and a cut library for six surfaces inside the same engagement.
1 hero · 22 cuts6 platformsSingle-vendor deliveryReference on request
◉ Marketing2024MarketingSaaS · Growth retainer · 2024Kavi SaaS
Role · Landing, lifecycle, paidThe problem.A B2B SaaS team had a credible product and an unfocused growth motion. Landing pages didn't differentiate by intent, lifecycle was a single welcome email, and paid was a residue of old experiments.
What we did.Full-funnel rebuild — intent-segmented landings, a five-step lifecycle, paid rebuilt from first principles. Trial-to-paid lifted; CAC drifted down through the quarter.
−31% steady-state CAC+58% trial-to-paidQuarterly cadenceReference on request