Koen Dercksen

Built a multi-tenant SaaS platform that tracks how AI assistants mention and recommend brands across all major LLM providers - visibility tracking for the age of AI answers.

• Scheduled prompt runs across 10+ AI engines with mention, ranking, competitor and citation analysis
• AI crawler and AI-referral analytics via a custom tracking script and CDN log ingestion
• Insights engine turning raw model output into scored, actionable recommendations
• Full agency multi-tenancy with subscription billing and an OAuth-secured MCP server

Built a sales-lead enrichment platform for an investment firm that qualifies companies by combining public registry data, procurement records, news signals and LLM web research into a single lead score.

• Pluggable connector pipeline over public data sources (business registries, insolvency records, procurement data, news)
• Background job queue that enriches and scores leads automatically
• LLM-powered web evaluation producing cited sales qualification briefs
• Review workflow and CRM export for the sales team

Automated the first assessment of car-lease applications for a lease provider, with a GDPR privacy-by-design architecture in which LLMs never process identifiable customer data.

• Pseudonymization vault with access-logged re-identification and an automated PII leak-scan gate
• Deterministic, auditable decision tree kept separate from LLM extraction
• SharePoint dossier integration and credit & business registry lookups
• Admin-editable policy engine with a full change audit trail

Built a custom chat environment with multiple LLM agents for a financial sector client, including a custom API backend.

• Developed custom LLM API layer in Python enabling observability and smart caching
• Configured and deployed multi-agent chat environment
• Integrated with client's existing infrastructure

Created an automated monitoring tool for a financial sector client that scrapes news nightly and detects relevant events regarding their client base.

• Automated nightly news scraping across multiple sources
• Built intelligent classification system for detecting insolvencies and noteworthy news
• Reduced manual monitoring time significantly
• Enabled early detection of client-related risks

Implemented multiple email generation and automation workflows for various clients using modern automation tools and LLMs.

• Created automated email workflows integrated with CRM systems
• Implemented LLM-powered personalized email generation
• Built scalable automation pipelines for marketing and customer communication

Developed a web-based viewer for simplified radiology reports as part of PhD research, making medical text more accessible to patients.

• Implemented automated medical text summarization
• Built entity annotation and recognition system
• Created entity description search functionality
• Designed patient-friendly interface for complex medical information