Available for new projects
$ ./lipht --status
# I help teams build solutions centered around AI and data — both hands-on and advisory.
get in touchservices.ts
function agenticEngineering() {
Bring AI and agentic coding into your team for real — hands-on training that goes past prompt tips, production-grade AI assistants built with proper evals and guardrails instead of brittle demos, and the open-source tooling to keep the habit alive long after the workshop ends.
}
function recommendationSystems() {
Design and build recommendation and personalization systems that hold up in production, not just in a notebook — cold-start handling, feedback loops, and the unglamorous plumbing (feature stores, offline/online parity, monitoring) that decides whether a model that scores well offline actually earns its keep in front of real users.
}
function dataPlatforms() {
Design and modernize analytical data platforms — infrastructure that serves machine learning and BI from one source of truth, integration pipelines that unify scattered source systems, and migrations off legacy stacks planned around the edge cases that actually break them, with compliance-sensitive data treated as a design constraint from day one.
}
work.log
$ lipht --deployments --status
AI coding harness — open source. Covers the full agentic dev lifecycle from idea capture through build, review, and self-improvement — 34 skills and counting, built and used in production, MIT licensed.
DevOps process design. Replaced ad-hoc deployment habits with a repeatable CI/CD pipeline on GitHub Enterprise, covering environment promotion and the guardrails that keep fast, frequent shipping safe.
Recommendation system on Databricks. Built from feature engineering through model serving, with a focus on holding up under real production traffic rather than just an offline evaluation metric.
Secure developer bootstrap platform. Gets new engineers and services to a working, compliant baseline fast — standardized tooling, access, and security posture from day one instead of every team reinventing it.
Claude Code adoption & agentic engineering training. Hands-on training for engineering teams on Claude Code and agentic coding — real codebases and real workflows, not a one-off demo, covering the skills and review habits that make agentic coding stick day to day.
Generalized AI interface for company workflows. Built on Azure with GDPR compliance as a first-class requirement, designed to be reused across internal use cases instead of shipped as a one-off point solution.
Payment data migration. Migrated sensitive payment data under GDPR constraints using a risk-based approach, sequencing and validating each move so nothing sensitive shifted without a clear compliance and rollback plan.
Advisory AI assistant. A retrieval-augmented assistant on Azure that grounds its answers in real source documents instead of relying on the model's own memory.
Salesforce Marketing Cloud implementation. Covered the underlying data model, personalized email content, and automated user journeys — turning raw customer data into targeted, triggered communication.
Matomo web tracking implementation. Gave the team first-party visibility into site behavior without handing that data to a third-party ad platform.
AI adoption & custom GPT training. Business-focused AI adoption training built around ChatGPT and custom GPTs — mapping AI into real business processes and helping non-technical teams build practical habits, well before agentic coding tools existed.
Recommendation system on SQL Server. Used collaborative filtering with explicit handling for cold-start users and items, instead of leaving new accounts with generic, one-size-fits-all recommendations.
Data platform for behavioral data. Captured web and email interactions and fed Power BI dashboards for reporting, with capture pipelines built to hold up under real production traffic.
Legacy CRM migration to Salesforce. Used a risk-based approach, sequencing the migration around the data and integrations most likely to break rather than a single big-bang cutover.
14 deployments. 0 failed.
about.md (preview)
# About
I'm Thomas Holk Nielsen, and I run Lipht Consulting. I work hands-on with teams on deploying solutions that utilize AI and data to capture value.
Impact first. Technology second. 10 years of experience working with companies in all sizes.
Strategic Advisor · Project Manager · Senior Developer · Architect
oss.log
$ lipht --open-source --status
AI coding harness for Claude Code — the full agentic dev lifecycle from idea capture to build, review, and self improvement. 34 skills that boost Claude Code performance for the pro dev, MIT licensed.
contact.sh
# Got a project in mind? I'd love to hear about it.
$ ./contact --email --phone