Senior Backend Engineer

Senior Backend Engineer.Saigon•On-site•Full-timeApply nowAbout Epsilo•We're building the operating system for performance media, unifying fragmented workflows across 50+ marketplaces globally•Trusted by the world's largest consumer brands including Unilever, P&G, L'Oréal, and Colgate•Backed by Sequoia, Vulcan Capital (founded by Microsoft co-founder Paul Allen), and K3 Ventures•Recognized by Gartner in their Market Guide for Digital Shelf Analytics for 3 consecutive years (2024, 2025, 2026)•Building from APAC to serve the fastest-growing performance media markets in the worldAbout the roleMarketplaces.

Social commerce. Quick commerce. Each retail media network has its own data model, its own vocabulary, and its own way of describing a campaign. A bid is not always a bid. A budget cap is not always the same field. An ad object on one platform may not have a direct equivalent on another.We built a Unified Campaign model to solve this — a single, normalized structure that abstracts away platform differences so that automation, analytics, and product features can be built once and work everywhere.As a Backend Engineer on the Data Unification team, you will own the pipelines and transformation logic that make this possible.

You'll work with raw performance media data, map it into clean, consistent structures, and ensure the unified model stays accurate and current, even as platforms evolve and edge cases multiply.This is a role for engineers who love the craft of data modeling and pipeline engineering, who take pride in making messy real-world data behave, and who want their work to be the foundation everything else is built on.The problem spaceEvery Performance Media Network exposes data differently.Campaign hierarchies differ: some platforms have ad groups between campaigns and ad objects; others don't.

Attribute names and semantics vary: daily budget, bidding strategy, status codes: none are standardized across platforms. Data freshness varies: some platforms push events; others require polling. Some fields update in real time; others lag by hours.Platform automations change state without notifying you — and your pipeline needs to detect and reconcile that.The Unified Campaign model is the layer that hides all of this complexity from the rest of the system.

Your job is to keep that layer correct, complete, and current.What you'll do•Build and maintain ingestion pipelines that pull campaign, ad group, ad object, and placement data from multiple marketplace APIs, handling polling, webhooks, event streams, and batch sync patterns•Design and implement the transformation layer that maps raw marketplace schemas into the Unified Campaign model, handling structural differences (platforms with/without ad group layers) and semantic differences (bidding strategies, budget types, status codes)•Ensure pipelines are resilient with retry logic, dead-letter handling, deduplication, and idempotent writes so partial failures never corrupt state•Implement reconciliation checks that compare unified model state against live marketplace state — surfacing drift before it becomes an incident•Build monitors around data freshness, completeness, and accuracy: field-level null rates, update lag, unexpected value distributions•Maintain the unified schema as new ad formats, platforms, and attributes are introduced — extending it without breaking existing consumers•Write clear mapping specs that document how each source field maps to the unified model, including edge cases and known platform quirksKey responsibilities•Own pipeline reliability: SLAs on data freshness, incident response, and continuous improvement•Collaborate closely with Distribution, Script Agent, Ads Operations, and Analytics teams.

Your unified model is the shared source of truth they all depend on•Participate in platform API reviews when marketplaces release changes: assess impact, plan migration, implement with zero downtime•Own RCA for data quality issues: trace discrepancies from the unified model back to the source event or transformation step•Document transformation logic and mapping decisions in living specs, not just code commentsWhat we're looking for•5+ years building data pipelines and ETL systems in production.

You have debugged schema mismatches, handled late-arriving data, and made messy third-party APIs behave•Strong data modeling instincts: you know when to normalize vs. denormalize, how to version schemas without breaking consumers, and how to design for extensibility•Experience with event-driven architectures (Kafka, Kinesis, Pub/Sub) and understand ordering guarantees, exactly-once vs. at-least-once semantics, and idempotency patterns•Comfortable working with SQL databases (PostgreSQL, MySQL, SingleStore) and know how to optimize queries, design indexes, and handle schema migrations at scale•Fluent with AI coding tools (Cursor, Claude, Codex) to accelerate development without compromising quality•Obsess over data quality: you build monitors before incidents happen, not after.

You know what good observability looks like for pipelines•Write clear specs and documentation. Your transformation logic should be understandable by the next engineer who inherits it•Clear technical communication and ability to work autonomously in a high-trust, low-process environmentOur stack•Frontend: Saasflow (proprietary low-code platform), React•Backend: Python, Java, Go, TypeScript, Node.js•Databases: Singlestore, Redis•Infrastructure: AWS, GCP, K8s, Kafka•AI-powered coding: Cursor, Codex, Claude, Graphite•AI collaboration: Notion, Linear, ChatGPTWhy join Epsilo•Own the foundation.

Every feature, every automation, every analytics surface depends on the unified model you build. Your pipelines are the shared source of truth for the entire platform•Solve real data engineering challenges. Schema evolution across 50+ marketplace APIs, late-arriving data reconciliation, platform quirks that can't be documented — this isn't Airflow on tutorial data•Work with world-class brands. Your transformation logic powers advertising operations for Unilever, P&G, L'Oréal, and Colgate across the fastest-growing performance media markets globally•Build for extensibility from day one.

Every new marketplace, every new ad format tests whether your abstractions hold. You'll design systems that scale both technically and conceptually•AI-native development environment. Use Cursor, Claude, and Codex to ship faster without compromising quality. Modern tooling for modern problems•Early team, massive leverage. Your architectural decisions around data modeling and pipeline design will shape how this platform scales to billions in ad spendHow we hire1Application reviewWe review your application, resume, and problem-solving note2Screening callCulture fit, role alignment, and answer your questions3Technical assessmentSystem design discussion and coding problem-solving4Team interviewDeep dive with engineers on architecture and collaboration5Founder conversationVision, values, and your questions about the company6DecisionWe move fastApply for this roleName *Email *LinkedIn URL Resume *Upload resumePDF only, up to 5MB.Show us how you operate *Three short answers.

Be specific and use numbers.How you 10x your output with AI: one workflow you turned into a reusable agent, skill, or Claude Code command. What was manual before, what the AI does now, and how much faster you are. *Link to that AI artifact (optional) A hard number you moved: a performance or reliability problem you owned. The metric, where it started, what you changed, where it landed, and how you kept it correct.

*A solution you designed end to end, from the user's real pain (not the tech): the insight, your approach, and what you deliberately chose not to build. *Submit application