PLANNING · 6 OCTOBER 2026 · AP-SOUTH-1
Managed AWS is the production direction for HirVault.
EvoHire owns the product. Telvrium may design, build and operate the technology. These figures are planning estimates for India, not an AWS quotation. GST, messaging, payment fees, model licences and salaries are excluded. Validate them with load testing and the AWS Pricing Calculator before procurement.
| Requests / day | Managed AWS | All EC2 | Planning choice | Stage |
|---|---|---|---|---|
| 1,000 | ₹35,000 – ₹70,000 | ₹36,000 | EC2 may be cheaper | Pilot |
| 10,000 | ₹45,000 – ₹90,000 | ₹50,000 | EC2 may be cheaper | Early production |
| 100,000 | ₹70,000 – ₹1.25L | ₹1.06L | Managed AWS | Growing |
| 500,000 | ₹1.15L – ₹2.10L | ₹1.83L | Managed AWS | Medium scale |
| 1,000,000 | ₹1.70L – ₹2.90L | ₹3.33L | Managed AWS | Large production |
| 2,000,000 | ₹2.70L – ₹4.30L | ₹5.15L | Managed AWS | Nationwide growth |
| 5,000,000 | ₹4.30L – ₹6.80L | ₹8.44L | Managed AWS | High scale India |
| 10,000,000 | ₹6.70L – ₹9.60L | ₹12.92L | Managed AWS | Large India-wide |
At about 10 million requests a day the managed range is ₹6.7L–₹9.6L a month, against about ₹12.9L for the all-EC2 model before staffing. Annualised, that is roughly ₹80.4L–₹1.15Cr versus about ₹1.55Cr. The EC2 design at that scale is about 45–55 nodes, not one server. Cognito cost follows monthly active users, and OpenSearch cost follows query complexity, so the managed range is not a simple sum of request charges.
What each method is responsible for
Method A · Managed
CloudFront and WAF, Cognito, ECS on Fargate, Aurora with RDS Proxy, OpenSearch Service, DynamoDB, ElastiCache, S3, SQS and EventBridge, SES, and CloudWatch. AWS carries more of the failover and patching.
Method B · Self-hosted on EC2
Nginx or HAProxy, Keycloak, PostgreSQL with Patroni, OpenSearch, Redis, MinIO, RabbitMQ, ClickHouse, and Prometheus, Grafana, Loki and Tempo. The team owns replication, recovery and the on-call.
Where the recommendation lands
- An internal pilot with a strong infrastructure team can use the EC2 stack.
- Commercial launch uses managed AWS.
- From about one million requests a day, and especially from five to ten million, managed AWS is the preferred path in this model.
- Choose EC2 when running the infrastructure itself is a capability Telvrium intends to staff.
Stack the product is designed around
- Public, recruiter, candidate and admin web
- Next.js, React, TypeScript, Tailwind
- Candidate mobile
- React Native, planned
- API
- Node.js and NestJS, REST first, UI calls the API only
- Authentication
- Amazon Cognito in production. This demo uses a local signed session.
- Authorisation
- HirVault RBAC inside the business services
- Business data
- Aurora PostgreSQL
- Search
- Amazon OpenSearch Service
- Inventory ledger
- DynamoDB
- Cache
- ElastiCache Redis or Valkey
- Files
- Amazon S3
- Async
- SQS and EventBridge
- Amazon SES
- Edge
- Route 53, CloudFront, AWS WAF
- Compute
- ECS on Fargate
- CRM
- Salesforce, for leads and the commercial lifecycle
- People system
- A separate HRMS, not the HirVault database
This running site follows the API-first rule in miniature: pages call route handlers, and those handlers are the only code that reads the local store. That store stands in for Aurora, OpenSearch, DynamoDB and S3 during development. It is not the production infrastructure.