About the role
Reports to: Head of AI & Digital
Type: Full-time
About the role
Apeiron collects and trades used cooking oil across Asia. Like many fast-growing companies, our data is fragmented across many systems and markets, including several ERPs and operational tools across Southeast Asia and China. Your job is to own the company's data foundation. You will engineer the pipelines that bring this data together reliably, model it into clean tables that finance and operations can trust, and keep it accurate and monitored. You will also prepare that data for AI, so that AI tools can search it and answer questions about our business correctly.
What you will do
- Build and run ingestion pipelines, starting with finance and ERP data, then operational sources
- Model the data into clear schemas and curated tables that finance and operations teams rely on
- Define and enforce data contracts, so it is clear who owns each dataset and who only feeds raw data
- Keep pipelines reliable: repeatable loads, late-arriving data handled, and failures that are visible and alerted
- Put quality checks where money breaks: amounts, dates, volumes, FX, and invoice against contract
- Own monitoring, lineage and access control for the data platform, from source system to report
- Prepare business data for AI ingestion so that AI assistants and search tools return accurate answers
- Work with finance and country operations in plain language so people can move off spreadsheets
Key skills we need
- Production data platforms: you have shipped ELT pipelines and production databases in messy, real companies, not only greenfield demos
- ERP and finance integration: Dynamics or a similar ERP; experience with AR, AP, bank and FX data matters
- Strong SQL plus Python (or Spark)
- PostgreSQL data platforms: hands-on with PostgreSQL-based platforms such as Supabase, or similar managed Postgres (Neon, Amazon RDS or Aurora Postgres), including pgvector for embeddings; comfortable moving data from raw to curated tables. Experience with cloud warehouses such as Snowflake is a plus
- Data contracts and reliability: schemas, late-arriving facts, monitoring, access control
- Hands-on: you build; you do not only draw architecture slides
- Cross-functional: you can sit with senior stakeholders and operations teams and explain things without jargon
AI data preparation (must have)
- Preparing data for AI ingestion: well versed in cleaning, normalising and chunking structured and unstructured data (tables, documents, emails, PDFs) so large language models (LLMs) can use it
- Vector embeddings and vector databases: hands-on experience generating embeddings and running a vector store such as pgvector on Postgres, Pinecone, Weaviate or similar
- Semantic search and retrieval (RAG): you have built retrieval pipelines, including hybrid keyword and vector search, so AI tools pull the right records before they answer
- Metadata and semantic layers: you can define business metrics, entities, relationships and rich metadata so AI tools and analysts query business data accurately and consistently
- Evaluation and freshness: you test retrieval quality and keep AI-facing data current, access-controlled and traceable back to its source
Nice to have
- Multi-country or supply-chain / commodity operations data
- Telematics or mobile collection feeds