MongoDB Architect / Developer
Key Responsibilities
Legacy Data Migration — DART Onboarding (10%)
A critical near-term initiative is migrating WON deals from the DART legacy quoting system into MongoDB Atlas. DART is one of four CPQ heads (alongside SAP CPQ, CDE, and MPC) feeding the D3 platform. This migration is complex for several reasons:
Data Source Challenges:
- Oracle DB origin: DART's master data resides in Oracle with a traditional relational schema — normalized tables for quotes, line items, pricing, customers, and approvals joined via foreign keys. This must be transformed into a single deeply nested MongoDB document per deal.
- XML export format: DART exports deal data as XML (nightly batch or on-demand trigger for WON quotes). The XML schema does not align 1:1 with the target JSON document model — requiring complex XSLT/programmatic field mapping across 160+ target fields.
- Incomplete data: DART XML exports lack several D3-required fields (SAP material numbers, contract terms, fulfillment details, billing schedules). These gaps are filled by calling real-time REST enrichment APIs during ingestion:
- TTSAPOUTPUTPKG API
- TTDOSEXTRACT API
Migration Pipeline: DART (Oracle) → XML Export → MuleSoft Validation → S3 Landing Zone → AWS Glue ETL (PySpark) + REST API Enrichment → MDS JSON → S3 Processed Zone → MongoDB Atlas
Your responsibilities:
- Design the MongoDB ingestion layer — bulk-write strategies, upsert logic (idempotent re-processing), write-concern tuning for batch loads, and temporary staging collections for validation before promoting to production.
- Build aggregation-based data-quality checks — validate that DART-sourced documents meet JSON Schema rules
- Implement a dead-letter collection for records that fail validation or API enrichment — with metadata capturing failure reason, retry count, and source XML reference for manual remediation.
- Optimize bulk-load performance — tune batch sizes, ordered vs. unordered inserts, index builds (background vs. foreground), and temporary index suppression during large migrations to maximize throughput.
- Track lineage — maintain full audit trail from DART Oracle source → S3 raw XML → enriched JSON → MongoDB document, enabling traceability for compliance and data-quality investigations.
Schema Design, Data Modeling and implementation (40%)
- Design and implement document schemas for Quote, Contract, Opportunity, and other entities that balance read performance with write efficiency — applying embedding vs. referencing trade-offs appropriate to each access pattern.
- Model deeply nested structures (5–6 levels) with careful attention to array growth patterns, document size limits (16 MB), and anti-patterns (unbounded arrays, deep-nesting without indexing).
- Implement multi-currency pricing using minor-unit integer representation across all Money-type fields; ensure aggregation pipelines handle currency conversion and precision correctly.
- Design and implement cross-collection reference strategies for Contracts → Quotes → Opportunities → Customers, using both manual references and $lookup where appropriate, and materialized views where join-heavy reads demand it.
- Define and maintain JSON Schema validation rules at the collection level to enforce required fields, data types, enum values, and structural constraints.
- Plan schema evolution — design migration strategies for backward-compatible schema changes as new entities and fields are added quarterly.
MongoDB Atlas Administration & Operations (5%)
- Provision and manage Atlas clusters on AWS — select appropriate tier (M10–M60+), configure auto-scaling, define shard keys for high-volume collections, and manage replica set topologies.
- Implement sharding strategies based on access patterns — compound shard keys (e.g., { CountryCode: 1, QuoteNumber: 1 }) for country-partitioned workloads, hashed shard keys for even distribution.
- Configure Atlas security — VPC peering / PrivateLink, IP access lists, SCRAM/x.509 authentication, field-level encryption (CSFLE/Queryable Encryption) for PII and financial data, audit logging, and RBAC with custom roles.
- Set up Atlas monitoring — Real-Time Performance Panel, custom alert conditions (oplog window, replication lag, slow queries > 100ms, connections, disk I/O), integration with CloudWatch and PagerDuty.
- Manage backup & disaster recovery — continuous backup with point-in-time restore (PITR), cross-region snapshot replication, and documented RTO/RPO targets.
- Optimize costs — right-size clusters based on working-set analysis, configure auto-scaling bounds, leverage reserved capacity, and analyze Atlas billing by cluster/namespace.
Indexing & Query Optimization (5%)
- Design compound indexes tailored to the platform's primary access patterns:
- Full deal retrieval: { QuoteNumber: 1, Version: -1 }
- Status-based listing: { Status: 1, CreatedTimestamp: -1 }
- Country-scoped queries: { "CountryDeals.CountryCode": 1 }
- Margin compliance: { "PortfolioOffers.Services.PerDevice.FloorMarginPercentage": 1 }
- Similar-deal lookup: { DealType: 1, "TCV.Total.amount": 1, LeadCountry: 1 }
- Build aggregation pipelines for real-time analytics — TCV roll-ups, cross-country pricing comparison, risk exposure summation, margin-drop analysis, service-module cost breakdown.
- Implement Atlas Search indexes for full-text search across deal descriptions, customer names, and service SKUs; configure analyzers, scoring, and facets.
- Configure Atlas Vector Search for AI-driven similar-deal lookup using pre-computed embeddings from the ML feature store.
- Profile and optimize using explain(), the Atlas Performance Advisor, slow-query log analysis, and index intersection strategies; eliminate collection scans and minimize index bloat.
- Design TTL indexes for ephemeral data (session state, draft quotes, AI inference cache) and time-series collections for audit event streams.
Integration & Pipeline Development (35%)
- Build change-stream consumers that propagate Quote/Contract mutations to downstream systems.
- Design and maintain Atlas App Services triggers (database triggers, scheduled triggers) for automated workflows: approval escalation, SLA breach detection, cost-model retraining kickoff.
- Integrate with the API layer — work with Go API developers to design efficient data access patterns, connection pooling (via the official Go driver), read/write concern tuning, and transaction boundaries for multi-document operations.
- Support the ML/AI pipeline — expose training data exports via aggregation pipelines or Atlas Data Federation; ensure the 180-day backfill window and 14-day retraining cadence are met without impacting production query performance.
- Configure Atlas Data Federation for cross-cluster and S3-archived data queries, enabling historical deal analysis without keeping cold data on hot storage.
Performance Engineering & Reliability (5%)
- Establish performance baselines — define SLIs/SLOs for read latency (P50/P95/P99), write latency, aggregation pipeline duration, and throughput (ops/sec) per collection.
- Load-test data-tier independently — simulate realistic deal-creation bursts, concurrent full-deal reads, and heavy aggregation workloads; validate auto-scaling behavior.
- Tune read/write concerns per operation criticality — majority for financial data, local for analytics reads, causal consistency for multi-step workflows.
- Implement connection management best practices — pool sizing, maxIdleTimeMS, retryable writes/reads, server selection timeout tuning.
- Conduct capacity planning — project storage growth, working-set size, and IOPS requirements for the next 12 months as Contracts, Opportunities, Customer and other entities come online.
Required Qualifications
MongoDB Certification (Mandatory)
Hold at least one active MongoDB certification:
Education & Experience
- Bachelor's degree in Computer Science, Data Engineering, or a related field — or equivalent practical experience.
- 5+ years of professional experience with MongoDB in production environments, with at least 2 years on MongoDB Atlas (cloud-managed).
- 3+ years designing schemas for complex, deeply nested document structures (not simple flat CRUD collections).
- Experience supporting enterprise-grade workloads — multi-tenant, multi-region, with strict latency and availability SLAs.