Forward Deployed Data Engineer

Job Title: Forward Deployed Data Engineer (Data Engineering & Big Data — Multi-Cloud, Client-Facing) QuanXAI | Bangkok, Thailand | Full-time, Hybrid (partial work from home)

About Us
QuanXAI is an AI technology firm delivering intelligent, scalable, and secure automation solutions for enterprises. Every AI solution we ship — agents, RAG systems, analytics, dashboards — stands on a data foundation that someone has to build inside the client's world: their clouds, their databases, their legacy systems, their security reviews. That someone is you. The Forward Deployed Data Engineer's goal is to turn messy enterprise data into AI-ready data products — at the client's site, on the client's stack. While AI Engineers focus on the Behavior (how agents think and models perform), you will own the Foundation and the Integration: discovering what data exists and where it hides, designing pipelines that fit the client's environment — whether that is AWS, Azure, or GCP — integrating with the systems they already run, and delivering production-grade data platforms that our AI products plug into. You are equal parts data engineer, system integrator, and trusted technical advisor. Job Responsibilities As a Forward Deployed Data Engineer, you will own the data side of client engagements end-to-end and take on the following: 1. Client Engagement & Data Discovery (25%)
• Discovery & Requirement Gathering: Run data discovery workshops with business and IT stakeholders — map source systems, data owners, refresh cadences, and quality issues; translate business questions (“why is churn rising?”) into concrete data requirements and quantified success criteria.
• Data Audit & Feasibility: Assess what the client actually has — profile databases, sample exports, inspect APIs — and give an honest verdict on what is feasible in 2 weeks, 2 months, or not at all with the data as it stands. Saying “the data can't support this yet, here's the path” is part of the job.
• Rapid Data POCs: Build credible proof-of-concept pipelines and datasets on real client data in 1–4 weeks — a working ingestion, a modeled dataset, a dashboard or an AI feature fed by it — designed from day one with a clear path to production, not a throwaway notebook.
• Communication & Demos: Present findings and demos in Thai and English to audiences from DBAs to executives — translating between “schema drift and CDC lag” and “why your report was wrong last Tuesday” fluently, managing expectations, and escalating risks early. 2. Data Pipelines & Big Data Engineering (30%)
• Batch & Incremental Pipelines: Design and build production ELT/ETL — ingestion from databases via CDC (Debezium, DMS, Datastream), APIs, SFTP and file drops; incremental processing with idempotency, late-arriving data handling, and backfill strategies that don't melt the source systems.
• Big Data Processing: Process data at scale with Spark (Databricks, EMR, Dataproc, Synapse/Fabric) and SQL-first engines — partitioning, broadcast vs. shuffle joins, skew handling, file-size and small-file management — knowing when a 2-node DuckDB job beats a 50-node cluster.
• Streaming Pipelines: Build real-time pipelines where the use case earns it — Kafka/Kinesis/Event Hubs/PubSub, stream processing (Flink/Spark Structured Streaming), windowing, exactly-once semantics, and dead-letter handling.
• Lakehouse & Warehouse Modeling: Design the storage and modeling layer per client — open table formats (Iceberg/Delta) with medallion architecture, or native warehouses (Redshift, BigQuery, Snowflake, Synapse) — with dbt transformation layers, dimensional models, and semantic layers that analysts and AI agents can both consume.
• Orchestration & Data Quality: Operate workflow orchestration (Airflow/Prefect/Dagster) with dependency-aware DAGs, freshness SLAs, and alerting; enforce data contracts and expectations (Great Expectations, dbt tests) at pipeline boundaries so bad data is quarantined, never silently propagated. 3. System Integration (SI) & Multi-Cloud Delivery (30%)
• Multi-Cloud Fluency: Deliver on whichever cloud the client runs — AWS (S3, Glue, Athena, Redshift, Kinesis, Lambda, DMS), Azure (ADLS, Data Factory, Synapse/Fabric, Event Hubs, Functions), GCP (GCS, BigQuery, Dataflow, Dataproc, Pub/Sub, Cloud Composer) — mapping the same architecture pattern onto each provider's services and knowing where the sharp edges differ.
• Enterprise System Integration: Integrate with what the enterprise already runs — ERP/CRM (SAP, Oracle, Salesforce, Dynamics), core databases (Oracle, SQL Server, PostgreSQL, MySQL), message buses, SSO/AD — through APIs, CDC, file interfaces, and vendor connectors, working within the client's change-management and security processes, not around them.
• Legacy & Constraint Engineering: Handle the reality of SI work — undocumented schemas, stored-procedure spaghetti, nightly batch windows, VPN-only access, approval-gated firewall rules — and still ship. Design around constraints instead of wishing them away.
• Hybrid & Cross-Cloud Architecture: Design data flows that span environments — on-prem to cloud, cloud to cloud — with secure connectivity (VPN, PrivateLink/Private Endpoints), encryption in transit and at rest, and network patterns that pass the client's security review the first time.
• AI Product Integration: Connect the client's data platform to QuanXAI's AI products — vector store ingestion for RAG (pgvector, Qdrant, cloud-native options), feature pipelines for ML, document processing pipelines — and expose governed datasets to agents via APIs and semantic layers. 4. Production Delivery & Handover (15%)
• Production Hardening: Take POCs to production — IaC (Terraform), CI/CD for data code (GitHub Actions/Azure DevOps), secrets management, monitoring and lineage (OpenLineage/dbt docs), cost controls, and access governance aligned with PDPA and the client's policies.
• Reliability in the Field: Define freshness/completeness SLAs, build alerting the client's team can act on, write runbooks, and support data incidents on systems you shipped — your pipeline, your pager, until handover completes.
• Handover & Enablement: Deliver documentation, architecture diagrams, and training so the client's team can operate what you built; leave behind an ROI story — faster reports, trusted data, AI features unlocked — that becomes the reference for the next engagement. Background / Experiences
• Engineering Foundation: B.
Eng. or B.
Sc. in Computer Engineering, Computer Science, or a related field — with strong fundamentals in data structures, algorithms, and distributed systems concepts (partitioning, consistency, idempotency).
• Production Data

Experience:
You have built and operated data pipelines that real consumers depended on — and debugged a 3 AM data incident, backfilled it, and added the test that catches it next time. A production data engineering stack is a requirement, not a bonus.
• SI / Consulting

Experience:
Experience in system integration, IT consulting, or client-facing delivery — you have worked inside someone else's environment, with their DBAs and security teams, and shipped despite the constraints.
• SQL & Python Depth: Expert-level SQL (window functions, query plans, optimization) and strong Python for data work — clean, typed, tested code, not throwaway notebooks.
• Client Communication: Proven ability to gather requirements from non-technical stakeholders, present technical trade-offs clearly, and build trust — in both Thai and English. Comfort being on-site at client offices as needed. Knowledge & Skills
• Multi-Cloud Data Stacks: Working knowledge of at least two of AWS / Azure / GCP data services (e.g., Glue+Athena+Redshift, Data Factory+Synapse/Fabric, Dataflow+BigQuery), with the ability to ramp on the third quickly.
• Big Data & Processing: Spark (Databricks or a cloud-managed runtime), dbt, and at least one streaming platform (Kafka, Kinesis, Event Hubs, or Pub/Sub); schema management (Avro/Protobuf).
• Databases & Warehouses: PostgreSQL plus at least one enterprise warehouse (BigQuery, Redshift, Snowflake, or Synapse); comfort reading unfamiliar schemas in Oracle/SQL Server; familiarity with vector stores (pgvector, Qdrant) for AI workloads.
• Orchestration & Quality: Airflow/Prefect/Dagster, Great Expectations or dbt tests, and CI/CD for data code; Docker, Git, and Terraform basics.
• Integration Patterns: REST/SOAP APIs, CDC, file-based interfaces, messaging, authentication flows (OAuth2, service principals, IAM roles), and secure network connectivity between environments. Bonus Points (Preferred if you have)
• Cloud Certifications: AWS Data Engineer / Solutions Architect, Azure Data Engineer (DP-203/Fabric), or GCP Professional Data Engineer.
• Enterprise Connector

Experience:
Hands-on integration with SAP (OData/IDoc/SLT), Salesforce, Oracle EBS, or Dynamics — or iPaaS tools (MuleSoft, Boomi, Fivetran, Airbyte).
• GenAI Data Pipelines: RAG ingestion at scale — document parsing (Docling/Unstructured), chunking/embedding experiments, vector index operations, or knowledge-graph construction (Neo4j).
• Streaming Depth: Flink or Spark Structured Streaming in production, exactly-once pipelines, or CDC with Debezium at scale.
• Analytics & BI: Semantic layer and dashboard delivery (Power BI, Looker, Superset, Grafana) — the last mile that makes data visible to clients.
• Craft Signals: Open-source contributions, data engineering write-ups, or side projects demonstrating pipeline craft and curiosity. QuanXAI — Making enterprise AI adoption seamless, secure, and measurable. Contact Nopawan@quanxai.com