Backend & Cloud Infrastructure Engineer

Mission

Build the data backbone, analytics engine, and SaaS platform that stores every inspection, computes every trend, and serves every dashboard. The SaaS model (₹3L/unit/year) depends on analytics delivering ongoing value – not just a one-time MR experience.

System Ownership

• Primary: Cloud API layer (REST + WebSocket APIs for all client-server communication)
• Primary: Inspection data storage (point clouds, deviation maps, QC results, audit trails)
• Primary: Multi-tenant SaaS architecture (workspace isolation, subscription management, enterprise SSO)
• Primary: Analytics data pipeline (raw inspection data → aggregated trend metrics)
• Secondary interface: Edge AI team (you provide ingestion APIs for edge-to-cloud sync)
• Secondary interface: MR team (you serve session state and multi-user anchor data)
• Secondary interface: Full Stack team (you provide APIs they build dashboards on)
• Does NOT own: ML model training (AI team), on-device inference (Edge AI team), MR rendering (MR team), front-end dashboards (Full Stack team)

What You Will Build

• Cloud inspection storage – Store LiDAR point clouds (100MB–2GB per scan), deviation maps, QC pass/fail results, and operator metadata. Handle ≥ 50GB new scan data per day across all tenants.
• Analytics engine – Deviation trend analysis over time. Which machines/structures show degrading quality? Which tolerance bands are consistently violated? Compute and cache aggregated metrics.
• Digital twin update pipeline – As new scans arrive, update the digital twin representation. Version both CAD revisions and scan history. Support rollback and historical comparison.
• Audit trail & compliance reporting – Generate tamper-evident inspection logs. Export as PDF and JSON for ISO/ASME compliance. Every QC decision must be traceable to a specific scan, alignment, and operator.
• Multi-tenant SaaS platform – Workspace isolation (no tenant can see another's data), role-based access, subscription tiers, usage metering, enterprise SSO (SAML/OIDC).
• Enterprise security – Data encryption at rest and in transit, API authentication (JWT + API keys), audit logging, SOC 2 readiness foundations, GDPR-aware data retention policies.

Core Technical Responsibilities

• Design and implement the REST + WebSocket API layer using FastAPI or equivalent – strict OpenAPI spec, versioned endpoints, rate limiting
• Build the point cloud storage layer: evaluate columnar storage (Parquet), spatial databases (PostGIS), and object storage (S3/MinIO) tradeoffs for query patterns vs. cost
• Implement the edge-to-cloud sync ingestion pipeline: receive compressed point clouds from edge devices, decompress, validate, store, trigger downstream processing
• Build the multi-tenant data isolation layer: schema-per-tenant vs. row-level security – choose based on tenant count and query patterns, implement and test isolation guarantees
• Implement the audit trail system: append-only log, cryptographic hash chain for tamper evidence, export to PDF/JSON with digital signatures
• Design the analytics aggregation pipeline: scheduled batch jobs + real-time incremental updates for deviation trend metrics
• Set up infrastructure: containerised microservices (Docker), orchestration (Kubernetes), infrastructure-as-code (Terraform), CI/CD pipelines

Required Technical Mastery

• API design: REST with OpenAPI, WebSocket for real-time updates, gRPC for internal service communication. Versioning, pagination, error handling standards
• Cloud platforms: AWS (primary) or GCP – production experience with EC2/ECS/EKS, S3, RDS, SQS/SNS, CloudFront, IAM, VPC networking
• Databases: PostgreSQL (primary), TimescaleDB or equivalent for time-series analytics, Redis for caching and session state, understanding of spatial query patterns (PostGIS)
• Microservices: Service decomposition, inter-service communication (sync + async), distributed tracing, circuit breakers, saga pattern for multi-service transactions
• Containerisation: Docker (multi-stage builds, security scanning), Kubernetes (deployments, services, ingress, HPA, resource limits, persistent volumes)
• Infrastructure as Code: Terraform or Pulumi – not ClickOps
• Security: OAuth 2.0 / OIDC, JWT management, SAML for enterprise SSO, encryption (AES-256 at rest, TLS 1.3 in transit), secrets management (Vault or AWS Secrets Manager)
• Languages: Python (primary – FastAPI, SQLAlchemy, Celery), Go (desirable for high-throughput services), SQL

Production Challenges You'll Solve

1. 50GB/day scan ingestion – 100 edge devices each uploading 500MB of compressed point cloud data daily. Your ingestion pipeline must decompress, validate, store, and trigger processing – without dropping data or blocking the edge device's sync queue.
2. Schema migration with production data – You need to add a new column to the inspection results table. 50 million rows. 12 tenants. Zero downtime. No data corruption. Build a migration strategy that handles this.
3. Tenant data leakage – A new developer writes a query that accidentally returns Tenant B's data to Tenant A's API call. Your isolation layer must make this architecturally impossible, not just "please be careful."
4. Audit trail tampering – A customer's compliance auditor asks: "prove this inspection result hasn't been modified since recording." Your append-only hash-chain audit log must answer this definitively.
5. Cost explosion – Uncompressed point cloud storage is growing at 1.5TB/month. At S3 standard pricing, this becomes unsustainable. Design a tiered storage strategy: hot (recent scans) → warm (last 90 days) → cold (archive), with transparent access across tiers.

Success KPIs

Failure If Underperforming

• API goes down → every edge device queues unsyncable data, every dashboard is blank, every MR multi-user session fails. Single point of failure for the entire platform.
• Tenant data leaks → immediate contract termination, potential legal liability, destroyed enterprise trust. One incident can kill the company at seed stage.
• Analytics are slow or wrong → the SaaS value proposition (ongoing insights, not just MR) collapses. Customers question why they're paying ₹3L/year/unit.
• Audit trail is tamperable → ISO/ASME compliance fails. Cannot sell to regulated industries (energy, aerospace, automotive). Addressable market shrinks by 70%.

Collaboration Interfaces

Why This Role Is Mission-Critical

Our SaaS revenue model depends on the backend delivering continuous value. The MR experience is the hook – analytics, compliance reporting, and digital twin updates are the retention engine. Without a reliable, secure, performant backend, every edge device is an isolated tool (not a platform), every customer churns after the novelty wears off, and the ₹3L/unit/year subscription cannot be justified.

About Us

Building the D2R (Design-to-Reality) platform – sub-millimetre CAD alignment + edge AI + mixed-reality overlay for industrial field workers. Venture-backed, seed-stage, < 20 engineers.

• Location: Bangalore / Hyderabad
• Stage: Seed / Pre-Series A (venture-backed)
• Industries: Construction, Manufacturing, Infrastructure, Energy