Backend & Data Platform Engineer

Backend & Data Platform Engineer

Company: Early-Stage Startup

Type: Full-Time Remote-Friendly

Level: Senior to Staff Engineer

About the Role

Building an ERP platform that autonomously runs the financial operations of small and mid-size businesses accounts payable, receivable, cash flow, procurement, and period close with humans approving only the decisions that require judgment.

You will be the first backend hire. You will design the data architecture, event-driven workflows, and API infrastructure that every AI agent in the platform depends on. This is not a feature-building role it is a foundation-building role. The decisions you make in the first 90 days will define the platform for years.

What You Will Own

Design the core multi-tenant data model entity graph, chart of accounts, transaction history, and workflow state with complete PostgreSQL Row-Level Security tenant isolation from day one

Implement event-driven architecture using Kafka/Redpanda topics, partitioning, consumer groups, dead letter queues, and exactly-once delivery for zero-loss financial event processing

Build high-performance async APIs in Python/FastAPI that AI agents call into tool endpoints, function schemas, context retrieval pipelines, and idempotent action handlers

Implement durable business workflows using Temporal.io 60-day collection cycles, PO approval chains, and month-end close sequences that survive server restarts and partial failures

Ensure ACID-safe financial transactions idempotency keys, saga patterns, distributed transaction recovery, and reconciliation logic for banking API integrations

Build the full observability stack structured logging, distributed tracing with correlation IDs, per-AI-call cost tracking, and alerting for every autonomous financial action

Own cloud infrastructure Kubernetes, Terraform, VPC design, IAM policies, and secrets management enforcing security at the infrastructure layer

Design the immutable AI audit ledger every agent action logged with full reasoning, reversibility metadata, and a read path that satisfies external auditors

Must Have

PostgreSQL (advanced) Row-Level Security, pgvector, TimescaleDB, schema design, indexing strategy, table partitioning, ACID transactions. You must be able to design a multi-tenant financial data model from scratch and defend every decision.

Event Streaming Kafka or Redpanda, event sourcing, CQRS, consumer groups, exactly-once delivery, dead letter queues, Avro/Schema Registry. Every AI agent action is event-driven this is core architecture, not plumbing.

Python / FastAPI Production-grade async Python, not notebook-quality code. SQLAlchemy (async), Pydantic v2, Alembic migrations, Pytest. You write the kind of Python that handles financial processes running without humans.

Durable Workflows Temporal.io (preferred), or demonstrable equivalent mental model via AWS Step Functions or Airflow. Long-running financial workflows must survive failures and time this is non-negotiable.

Strong Plus

Cloud infrastructure (AWS or GCP) Terraform, Kubernetes, IAM, Secrets Manager, VPC design

Observability OpenTelemetry, DataDog/Grafana, structured logging, distributed tracing, SLOs

AI agent backend integration LLM tool/function schemas, pgvector RAG pipelines, structured output contracts, context window management

Financial API integration Plaid/Teller, Stripe/Adyen, webhook handling, idempotency, retry/backoff, circuit breakers

Tech Stack

PostgreSQL pgvector TimescaleDB Redis Redpanda/Kafka Temporal.io Python 3.12 FastAPI SQLAlchemy Pydantic v2 Anthropic Claude API LangGraph AWS/GCP Kubernetes Terraform OpenTelemetry DataDog