Scaled sandbox testing from 28 to 1,200 tests monthly with Google Kubernetes Engine (GKE)
Cut driver onboarding times from days to minutes with Gemini
Achieved 30% zero-touch driver verification automation on Gemini Enterprise Agent Platform
Slashed document processing costs by 75% using Gemini Flash
Reduced Code-to-Deployment lead time down to ~2 days on Google Cloud
Porter leverages GKE and Gemini 2.5 Flash on Google Cloud to modernize microservices and automate driver onboarding.
Porter leverages GKE and Gemini 2.5 Flash on Google Cloud to modernize microservices and automate driver onboarding.
Overcoming cross-border latency and legacy operational bottlenecks
Porter (Smartshift Logistics Private Limited), a leading logistics platform based in India, orchestrates on-demand mini-trucks, tempos, and two-wheelers across 50+ major cities. When order volumes doubled, Porter’s platform engineering team began facing severe scaling constraints on legacy infrastructure.
Core compute was historically hosted in Singapore, introducing cross-border network latency for API calls originating from drivers and merchants across India. Infrastructure management was centralized, creating operational bottlenecks and restricting deployment autonomy for individual application microservice teams.
At the same time, infrastructure spend was not aligned with business scale, and the company needed to strengthen its security posture. Onboarding thousands of new driver-partners daily presented another major operational hurdle. Verification relied on a third-party OCR vendor to extract identity data from Aadhaar, PAN, driving licences, and vehicle registration certificates (RC). The vendor charged high unit fees (~1 Rupee per document), exhibited high failure rates requiring manual retries, and offered only 75% first-attempt accuracy. This forced every application into prolonged manual review queues, stretching driver onboarding turnaround times across several days and causing candidate drop-offs during high-demand surges.
We wanted to optimize our cloud costs while modernizing our infrastructure with Kubernetes. Moving to GKE gave us the right set of tools to manage our workloads with minimal burden, improve our security posture, and eliminate operational bottlenecks for our engineering teams.
Bijoy Paul
Principal Engineer, Porter
Porter migrated its compute and storage footprint to Google Cloud in order to eliminate cross-border network overhead, optimize infrastructure spend, and modernize its application stack. With Google Kubernetes Engine (GKE) and Gemini Enterprise Agent Platform, Porter established a resilient, developer-centric foundation designed for high-concurrency real-time logistics.
Modernizing microservices and automating verification workflows with GKE and Gemini
Porter restructured its core application runtime on GKE, standardizing deployment pipelines using GitOps with Argo CD and Argo Rollouts. To decentralize traffic governance, the engineering team adopted GKE Gateway API to modernize its Kong Gateway architecture. To solve DNS lookup bottlenecks during rapid autoscaling bursts, Porter deployed NodeLocal DNS Cache across its clusters.
For data management, Porter migrated 130+ PostgreSQL databases to Cloud SQL for PostgreSQL, leveraging PostGIS and Cloud Pub/Sub to process real-time geospatial telemetry pings from active drivers. To handle 0.5 Petabytes (500 TB) of monthly application logs economically, Porter deployed a self-hosted Vector collector on GKE streaming directly to Cloud Storage, visualized seamlessly using Grafana. Meanwhile, to boost developer velocity, Porter built “Flash”, an in-house sandbox framework on GKE and Cloud Pub/Sub allowing engineers to run 1,200 parallel feature tests monthly.
With the adoption of Gemini 2.5 Flash on Gemini Enterprise Agent Platform, we brought OCR directly under our control. It provided superior extraction accuracy compared to our previous vendor at a fourth of the cost, with significantly lower processing latency and zero retry overhead.
Upendra Datt
Principal Engineer, Porter
To transform driver onboarding, Porter deployed Gemini 2.5 Flash on Gemini Enterprise Agent Platform to extract structured JSON data from identity documents. To comply with India's DPDP (Digital Personal Data Protection) and Aadhaar Acts, Porter relies on Cloud Key Management (KMS) for symmetric key encryption. This delivers a consistent encryption standard, straightforward key rotation, and clear, centralized logs of who decrypted what—across data at rest (databases, Cloud Storage, backups) and at the application layer. Porter also recently adopted application-layer protection, finding Cloud KMS to be the most cost-effective and operationally simple way to encrypt and de-identify sensitive PII before it is written to storage.
Finally, the team built an LLM flow on Gemini Enterprise Agent Platform that classifies vehicle make/model data into logistical payload types with 97.5% accuracy. Porter also developed “Terraform Sensei”, an internal multi-agent assistant built on Google Agent Development Kit (ADK) that captures Terraform workflow failures and posts automated root-cause fixes directly as GitHub pull request comments.
Seamless policy approvals optimize team efficiency and customer care
Migrating to Google Cloud transformed Porter’s platform velocity, operational efficiency, and driver onboarding capability across all urban hubs. By replacing full-environment clones with the ‘Flash’ testing sandbox, the platform team expanded parallel testing capacity from 28 to 1,200 isolated feature tests per month, driving Code-to-Deployment (C2D) time down to approximately two days. Concurrently, adopting GKE blue-green upgrades eliminated maintenance downtime during control plane and node pool updates, ensuring continuous uptime during high-demand booking surges.
The integration of Gemini 2.5 Flash fundamentally restructured Porter’s onboarding unit economics and partner conversion rates. Replacing legacy OCR vendors with Gemini reduced document processing costs by more than 75%, dropping unit costs from one rupee down to roughly 25 paise per document while improving first-pass extraction accuracy from 75% to 91%. Processing latency fell by 50%, enabling 30% of new driver onboardings to be executed automatically with zero manual intervention, and slashing verification turnaround times from several days to just minutes.
Automated vehicle categorization on Gemini Enterprise Agent Platform achieved 97.5% classification accuracy, eliminating manual back-office sorting for complex three-wheeler and truck make/models.
Automating document verification and vehicle categorization removed the human out of the loop for 30% of onboardings. Slashing turnaround times to minutes gives us a massive win in driver conversion and operational efficiency across 50+ cities.
Upendra Datt
Principal Engineer, Porter
Looking ahead, Porter is benchmarking Gemini 3.1 Flash models to further reduce document extraction costs to 10 paise per document. The engineering team is also actively preparing to expand Gemini-based agentic workflows across platform operations and support systems, while evaluating BigQuery to drive the company’s next-generation real-time Customer Data Platform.
Porter is a leading Indian logistics platform offering on-demand intra-city freight, mini-truck, and courier services across 50+ cities, connecting millions of users with driver-partners.
Industry: Transportation and Logistics
Location: India
Products: Google Kubernetes Engine (GKE), Gemini Enterprise Agent Platform (formerly Vertex AI), Gemini 2.5 Flash, Cloud SQL for PostgreSQL, Cloud Storage, Cloud Pub/Sub, Cloud Key Management Service (KMS), Artifact Registry, Cloud Run, Cloud Composer, BigQuery
