Jenesys AI – Data & AI Engineering for an AI Bookkeeper
A senior engineer embedded with Jenesys AI, a London fintech building Jack, an AI co-pilot for bookkeeping, working between its AI and core engineering teams.
- Client
- Jenesys AI (London, UK)
- Industry
- Fintech / AI accounting
- Period
- Jul 2024 – Mar 2025
- Engagement
- Embedded senior AI & data engineer
Key Results
Extraction-quality alerts for key bookkeeping fields and a feedback loop from customer corrections into the model's embeddings store
The Challenge
Jack reads invoices and receipts and proposes the bookkeeping entry: category, VAT rate, tracking categories. After its $1.1M pre-seed round, Jenesys needed an engineer who could work across the AI team and the Python/AWS engineering team to make extraction quality measurable and feed customer corrections back into the models.
Our Solution
Our senior engineer joined the team in July 2024, working in Python on AWS. In the orchestration service he built extraction-comparison checks with alerts for missing category, VAT-percentage and tracking-category values, and a pipeline that pushes AI feedback into the orchestrator queue so it is indexed into Jenesys's ClickHouse-backed embeddings store. He also worked on logging for in-context learning from client feedback, and on feature-engineering and data-pipeline support for the AI team.
The Outcome
Missing values in key extraction fields now raise alerts instead of passing silently, and customer corrections flow back into the store the models learn from. The engagement started through Smart Maple in July 2024 and continued as a direct consultancy into 2025.
Highlights
- Extraction comparison with null-value alerts
- Customer feedback indexed into a ClickHouse embeddings store
- In-context-learning logging from client corrections
- Feature engineering and pipelines for the AI team
Related Projects
EdgeHunters – Real-Time Odds Data Platform
Scrapers, normalization and quality checks that feed live odds from dozens of sportsbooks into a betting-analytics product for US and Canadian users.
Tribute Technology – Funeral Home Data Migration Platform
A self-service platform that converts funeral homes' legacy data into Tribute's format, so new customers can move to CRäKN (now Tribute Management Software) without one-off manual imports.
AI Exercise Program Generator for a Sports Physiotherapy Clinic
A web app that drafts individualized strength and rehab programs from an athlete's test data, so physiotherapists start from a sound plan instead of a blank page.
Interested in Data & AI Services?
Learn more about our data & ai capabilities and see how we can help with your project.
Explore Data & AI ServicesHave a Project in Mind?
Let's discuss how we can bring your vision to life with a custom solution.
Start Your Project