Tripeur is an end-to-end corporate travel management platform that uses advanced Artificial Intelligence and Machine Learning algorithms to drive efficiency, ease of use, personalization and transparency for business travel by employees and executives alike. This project sits slightly outside travel booking and squarely inside a customer's finance function, addressing GST tax reconciliation for travel invoices across companies, airlines, hotels, railways, and bus vendors.
GST arrived in 2017 promising one nation, one tax, yet each year companies lose hundreds of crores in credit through lack of proper reconciliation among multiple parties. The word that matters is reconciliation rather than filing. The tax has been paid and the credit is legitimately claimable. It goes unclaimed because proving entitlement means matching records nobody has time to match.
GST aimed to simplify taxation under one nation, one tax, yet companies lose hundreds of crores in credit annually through poor reconciliation among multiple parties. The tax has already been paid and the credit is legitimately claimable. It goes unclaimed because proving entitlement requires matching a company's records against what each vendor separately reported.
Travel invoices arrive in various formats from airlines, hotels, railways, and bus vendors. Travel is unusually fragmented here, since one business trip can generate invoices from four different vendor types. Unlike a supplier relationship where format can be negotiated, a travel vendor issues whatever it issues and the recipient has no leverage.
On average, companies receive a staggering 150,000 invoices annually, and manually reconciling tax on each would be a logistical nightmare for any finance department. The arithmetic settles it: even one minute per invoice is 2,500 hours, more than a full-time year spent purely matching documents. No team is staffed for that.
The volume led many larger companies to forego claiming their GST input credits altogether. This is counter-intuitive until the mechanics are clear, since a larger company generates more invoices, so its unclaimed credit is bigger and the manual effort to claim it is proportionally bigger. Those with most to gain found it most impossible.
Automating only the matching would have failed, because matching is not where the work is. Collecting invoices from many vendors in many formats is the bulk of it. Worse, you cannot reconcile an invoice you never received, and a missing invoice is invisible by definition. Tripeur therefore scoped a product automating the entire process end to end, covering ingestion, parsing, comparison, and reporting.
Separate email addresses were set up per Tripeur customer and integrated with Freshdesk, so a vendor email landing there was ingested and stored securely in S3. Vendors already send invoices by email, so onboarding becomes redirecting an address rather than changing how a finance team works, and every invoice arrives with a ticket trail behind it.
A parser service handled PDF, image, and text invoices, transforming the data into a user-friendly format. Supporting all three is what lets the system accept whatever a vendor sends rather than only what it prefers. A parser covering only the cleanest format pushes everything else back to manual handling, restoring the bottleneck it was built to remove.
A service identified missing invoices and retrieved them from the vendor portal or sent periodic email reminders. This addresses what nobody else can see, since reconciling what you hold says nothing about credit you never received documentation for. Knowing from booking records that an invoice should exist is what makes the absent visible.
The reconciliation system compared data and highlighted missing invoices, tax amount discrepancies, and unclaimed credit amounts. Those three outputs match three different failure modes needing three different responses: chase the vendor, query the figure, file the claim. A single total would say something is wrong without saying what to do about it.
The product enabled companies to claim approximately 1 million dollars in credit. This is recovered money rather than saved cost, since the tax was already paid and simply went unreclaimed.
The product helped companies find approximately 10,000 missing invoices, each representing credit that was entitled, undocumented, and invisible to any process working only from invoices already held.
The successful product launch resulted in a 6 percent expansion of Tripeur's customer base, reaching buyers travel management alone would not have won, because it speaks to finance rather than travel.
Eternalight built Tripeur's GST reconciliation product on Node.js and Python with an AngularJS frontend, running serverless with a message queue coordinating the pipeline. Invoices arrive through dedicated per-customer email addresses integrated with Freshdesk and are stored in AWS S3, then handled by parser services covering PDF, image, and text before reconciliation and reporting.
Brainstorming
Wireframe
Development
User Feedback
Launch
Integration
A cross‑platform, open‑source JavaScript runtime environment for building fast, scalable server‑side applications.
Powers backend logic and AI-driven processing.
Angular: A TypeScript‑based framework for scalable, enterprise single‑page applications.
A comprehensive cloud platform offering compute, storage, databases, and AI services at global scale.
Scalable processing without managing infrastructure.
Coordinates invoice processing across multiple services.
Send us a message, and we’ll promptly discuss your project with you.
