Data Architect

Req: Data Architect with hands on exp in AWS Bedrock, ETL, SQL, Informatica

Location: Remote

Key Skills Required

Must-have (prioritize candidates with depth in these areas):

Priority Skills

Critical Amazon Bedrock model selection, prompt design, Knowledge Bases, retrieval-augmented generation (RAG), guardrails, and production inference patterns

Critical AWS platform CI/CD (CodePipeline / GitHub Actions on AWS), Lambda, Step Functions, OpenSearch (vector search), Neptune or graph storage, Secrets Manager, S3 artifact pipelines, IAM, and cost/scale design

Required SQL Server at scale stored procedure dependency tracing, cross-database schema reconciliation, RLS / SESSION_CONTEXT

Required Dispositioning & triage turning automated outputs (schema drift, source-only tables, static-analysis findings) into auditable, dependency-backed decisions

Required ID de-collision analysis overlapping ID ranges, true collisions vs. co-mingled data, table-level resolution

Required ETL / batch consolidation Informatica (or equivalent), migration script quality, tenant-scoped batch jobs

Required LLM output validation guardian / evaluation patterns, golden-set regression, confidence thresholds

Required Code literacy (.NET, Java, Angular, Python); static analysis & SARIF; financial-services consolidation experience

---

Role Summary

Senior practitioner who combines deterministic data engineering with Bedrock-powered AI reasoning to accelerate a large-scale multi-tenant consolidation. You turn automated signals schema drift, source-only tables, ID collisions, and static-analysis findings into defensible disposition decisions grounded in dependency chains (stored procedures, application code, ETL, batch jobs). You also productionize the toolchain on AWS: RAG through CI/CD, knowledge capture, and a guardian pattern that validates every model output before it reaches reviewers or migration teams.

---

Core Responsibilities

1. Dispositioning Reasoning

Adjudicate schema drift, source-only tables, and migration-priority findings using SP app ETL dependency evidence.

Apply deterministic rules where the answer is deterministic; invoke Bedrock RAG where genuine judgment is required.

Produce auditable dispositions with confidence levels and blast-radius assessment; feed validated decisions back into the knowledge base.

2. ID De-Collision Reasoning

Investigate overlapping ID ranges; distinguish true collisions, co-mingled data, and namespace overlap without semantic conflict.

Define table-level resolutions (re-keying, offset mapping, surrogate keys, RLS-scoped acceptance) with end-to-end traceability through SPs, apps, and ETL.

3. Data Migration & Batch Consolidation

Review migration script quality (idempotency, ordering, rollback, tenant scoping).

Consolidate Informatica / batch jobs for tenant-specific routing; align with comment-out vs. refactor vs. data-layer-only strategies.

Support cutover sequencing, reconciliation, and post-migration validation.

4. Productionizing the Toolchain on AWS

Integrate the Bedrock RAG disposition assistant into CI/CD (retrieval over dependency graphs, schema diffs, SME feedback, prior dispositions).

Implement a guardian / evaluation pattern: grounding checks, structured-output validation, confidence thresholds, escalation for high-blast-radius decisions, golden-set regression.

Operationalize SME knowledge capture, observability, and versioning for prompts, indexes, and disposition rules.

5. Toolchain Assessment

Outside-in review of the code-graph generator (resolution accuracy, build vs. leverage, polyglot coverage).

Recommend AWS scaling paths: Bedrock Knowledge Bases, OpenSearch vectors, Neptune, serverless batch, SARIF/report pipelines.

Deliver a prioritized roadmap with measurable coverage improvement (EAC and disposition throughput).

---

Qualifications

Required

8+ years in data engineering, platform migration, or enterprise modernization (financial services preferred).

Production experience with Amazon Bedrock and core AWS services (not exploratory POCs).

Large-scale SQL Server dependency analysis and schema reconciliation.

Hands-on ETL/batch consolidation (Informatica or equivalent).

CI/CD integration and LLM output validation in production workflows.

Preferred

Multi-tenant consolidation, tenant discriminators, feature entitlements.

Code graph / static analysis (Roslyn, OpenRewrite, CodeQL, custom analyzers).

Guardian patterns for LLM outputs; Python/PowerShell report pipeline automation.

---

Engagement & Success Metrics

Phase Focus

Assess Toolchain review; disposition backlog; ID collision inventory

Establish Disposition standards; Bedrock RAG + guardian pattern; CI/CD skeleton

Execute Domain dispositioning; migration/batch review; ID resolution

Operationalize Knowledge loops, metrics, program handoff

Success: 90% SME acceptance on first review; guardian catches 95% of ungrounded model outputs; zero post-cutover ID collision defects in pilot domains; documented AWS toolchain roadmap with coverage targets.