10/12/2026 updated
100 % available
AI Engineer | LLM Evaluation, Reliability & RAG | Python Automation
Delft, Netherlands
Only remote
Master of Business Administration (MBA), Data Analytics SpecializationAbout me
I build and troubleshoot AI systems, with a focus on reliable outputs, measurable quality, and clear handoffs. I bring 10+ years of software and automation experience and take ownership from the initial problem through implementation and evaluation. Based in the Netherlands; working remotely.
AI AgentsRAGLLMOpenAIGCPLlamaIndexLangChainOllamaArtificial IntelligenceElasticsearchGitHubInformation RetrievalPython (Programming Language)Machine LearningSQL Databases
I build and improve AI applications that need reliable retrieval, validated outputs, and measurable quality. I work across RAG, LLM evaluation, agentic workflows, and Python/API integration, taking ownership from diagnosing the problem through implementation, testing, monitoring, and handoff.
RAG & DOCUMENT PROCESSING
Build document ingestion and retrieval workflows using LlamaIndex, LlamaParse, embeddings, and semantic search. Diagnose retrieval failures, evaluate chunking and context choices, and test whether changes improve performance across representative questions.
LLM EVALUATION & RELIABILITY
Build evaluation suites, categorize failures, and validate changes to prompts, retrieval, models, and workflow logic. Add deterministic checks around model outputs using Pydantic, JSON Schema, and content validation. Production experience maintaining an approximately 40-case evaluation suite; public projects demonstrate retrieval benchmarking and structured-output checks.
AI WORKFLOWS & INTEGRATION
Connect LLMs to business systems through custom Python, REST APIs, and LangGraph orchestration (initial prototype in Dify). Combine retrieval, model calls, structured outputs, validation, and downstream automation. Architected and deployed a six-stage production AI workflow using multiple model providers.
DEBUGGING & PERFORMANCE
Investigate failures across inputs, schemas, retrieval, model responses, and downstream systems. Use evaluation results to guide model selection and prompt/context changes. Reduced production AI latency by approximately 20% and token usage / per-request cost by approximately 10%.
PYTHON & BUSINESS AUTOMATION
Build and repair API integrations, ETL pipelines, data transformations, and recurring workflows with validation, retries, deduplication, logging, and monitoring. At Smarsh, reduced ETL-related failures by 60% and eliminated 1,000+ redundant API calls per day.
TECHNICAL STACK
Python, SQL, OpenAI, Anthropic, Gemini, LlamaIndex, LlamaParse, Dify, Pydantic, JSON Schema, embeddings, REST APIs, Git/GitHub, Docker, Datadog, AWS, Azure, and Azure Data Factory.
Keywords: KI-Entwickler, KI-Engineer, Generative KI, LLM-Entwicklung, RAG
10+ years of software and automation experience. Netherlands-based, native English speaker, available for remote B2B engagements.
Public code and evaluation examples: github.com/regankight
Have an LLM or RAG workflow that needs building or debugging? Send me the scope, current bottleneck, and timeline through Freelancermap.
Experience
Professional experience10+ years
Freelancer since2026
Languages
EnglishNative speaker
Project history
Designed and deployed a 6-stage production RAG/LLM workflow transforming unstructured stakeholder input into implementation-ready technical specifications. Built a custom engineering layer in Python with API integrations, Pydantic/JSON schema validation, data transformations, retrieval configuration, evaluation code, logging, and monitoring. Integrated OpenAI, Anthropic, and Gemini models. Established an LLM evaluation framework with approximately 40 test cases. Built multi-step agentic workflows using Dify orchestration. Optimized token usage and latency by approximately 10% and 20% respectively.
Designed and hardened Python ETL and API integration pipelines across Salesforce, Marketo, HubSpot, and Power BI, reducing ETL-related failures 60%. Diagnosed production failures through repeatable triage sequences. Rebuilt the Salesforce to Power BI integration eliminating 1,000+ redundant API calls per day. Implemented Talend fallback logic and pre-aggregation cutting pipeline lag 70% and preventing $100K+ in revenue forecasting errors. Built automated anomaly detection for production data.
Delivered enterprise account intelligence for Amazon, Google, and Peugeot, combining internal revenue and sales-funnel data with external market intelligence. Extracted and transformed enterprise data using SQL and Python, built a reusable analysis framework, and developed a USPS API geolocation-cleanup pipeline that reduced manual effort 60%.