Senior data/backend engineer to productionise an existing AI extraction pipeline

We’re building Mineral Metrics, a data platform for the mining industry. Our existing system collects company announcements and uses AI to extract structured financial, operational and project data. A substantial amount of the extraction pipeline has already been built and is working. The challenge is turning its output into a reliable, maintainable database that can be consumed by our existing dashboard application. We are looking for a senior data/backend engineer with practical experience building production data pipelines involving LLM extraction. This is not a greenfield rebuild. Our preference is to preserve and improve the existing system, understand the decisions already made and find the shortest path to getting our near-complete MVP into production. INITIAL SCOPE The first engagement will be a paid technical audit. You will: - Review the existing codebase and current data flow - Understand how source announcements are collected and processed - Assess the extraction, validation and provenance logic - Review the current output structure and dashboard requirements - Identify reliability, maintainability and scaling risks - Recommend a practical database and integration architecture - Separate what should be retained, repaired or replaced - Produce a prioritised plan to reach a working MVP If the audit goes well, we would like the same person to help implement the plan alongside our technical co-founder. The eventual implementation may include: - Designing or improving the database schema - Normalising extracted data without losing source provenance - Handling reprocessing, duplicates and revised announcements - Creating stable interfaces between the extraction pipeline and dashboard - Adding validation, error handling and monitoring - Supporting historical backfills - Reducing manual intervention while preserving accuracy WHAT MATTERS TO US - Strong Python and/or TypeScript backend experience - Production experience with PostgreSQL or a comparable relational database - Experience with ETL/ELT pipelines and data modelling - Experience turning LLM output into validated structured data - Familiarity with provenance, audit trails and versioned data - Ability to work with an existing codebase without immediately proposing a rebuild - Clear communication with a capable technical founder - A practical MVP mindset Experience with mining or financial data is helpful but not required. Our priorities are accuracy, traceability and getting the existing product into users’ hands. PLEASE ANSWER THESE QUESTIONS IN YOUR PROPOSAL 1. Describe one production system where you converted documents or unstructured text into structured database records. What did you personally build? 2. How would you preserve the source and provenance of an AI-extracted number so a user could trace it back to the original document? 3. How would you handle an announcement being processed twice, corrected later or reprocessed using a newer extraction model? 4. Tell us about an existing system you improved without rebuilding it. What did you retain and what did you change? 5. What would you inspect during the first few hours of this engagement? Please do not send a generic list of AI tools. We are more interested in systems you have shipped and the engineering decisions behind them. INITIAL ENGAGEMENT Paid fixed-price technical audit expected to require approximately 5–8 hours. Deliverables: - Short architecture and data-flow assessment - Key risks and gaps - Recommended target architecture - Prioritised MVP implementation plan - Rough implementation estimate - One review call with the founders There is follow-on implementation work available if the initial engagement is successful.