Internship - Full-Stack / AI Engineer

About Finamina Finamina is building the cash layer for businesses in MENA : a system that reads a company's financial documents, works out what is owed and by whom, reconciles it against the bank, and collects, without anyone opening a spreadsheet. For an engineer, that requires turning unstructured paper (ie. invoices, purchase orders, delivery notes, arriving as scans and email attachments in three languages) into a reliable, real-time picture of a company's cash position. The documents are messy, the formats never agree, and on financial data a plausible-looking wrong answer is worse than no answer at all. The product is live in production with an ambitious roadmap. The role The work splits between consolidating what already exists and building new capability: primarily AI agents that execute business workflows on behalf of the customer, and strengthening the scoring layer. On working

conditions:
Finamina is early stage. Priorities are re-arbitrated against product and commercial urgencies, and the scope of this role will not stay fixed across the internship. The product backlog is the working baseline, but execution order shifts. Responsibilities Product consolidation Hardening existing ingestion and extraction pipelines, reducing error rates on degraded documents, building and strengthening third-party integrations. AI agents and workflow automation Designing and implementing agents that carry out business tasks in-app on the customer's behalf: classification, follow-ups, reconciliation preparation, conditional actions. Defining guardrails and human validation checkpoints. Scoring intelligence Extending the product's scoring layer: modelling, input signals, and how results surface in the interface. Product development Backend and frontend work across the backlog: API, review and validation interfaces, integrations. Technical environment Python 3.11 and FastAPI (fully async) on managed PostgreSQL, Next.js on the frontend, and an LLM-based document extraction stack combining hosted inference with OCR. CI/CD through GitHub Actions across separate staging and production environments. The product is API-first: a fully asynchronous REST API with strict schema validation, designed to sit between a company's documents, its bank, and its accounting system. A significant share of the engineering work is API design: both the interface the product exposes and the third-party APIs it consumes. Requirements
- Engineering degree or Master's in computer science, data or equivalent (final or penultimate year)
- Python and web development (React / Next.js or equivalent)
- Hands-on experience applying LLMs in production contexts: structured output, pipeline integration, quality evaluation
- Professional working English
- Able to make progress without a closed scope and to absorb shifting priorities
- Self-directed in a remote setting, without structured daily supervision Terms
- Duration 4 to 6 months, adjustable to your school's agreement
- Start September 2026, exact date to be agreed
- Location Fully remote
- Commitment Full time