AI that delivers value.
And survives production.
I help teams decide where AI pays off, then build and operate the predictive models, workflows and agents that deliver it.
Led by Stefan Jansen since 2015.
Areas of work
-
AI strategy and capability development
Make informed decisions about where AI belongs, which projects to develop and what your organization needs to implement it. I work with leaders and technical teams on priorities, learning and practical choices.
Prioritization and investment cases, feasibility and data readiness, executive education
-
Predictive analytics and decision science
Build forecasts and customer models that support planning, pricing and commercial decisions. I develop the analysis, models and operating systems that put those decisions into practice.
Demand forecasting, churn and uplift modeling, causal inference, pricing and targeting decisions
-
AI agents and workflow automation
Automate the parts of a business process that software can handle well, and give people the information and control they need for the rest. I design and build document systems, AI-assisted workflows and agents.
Contract intelligence and document review, agent workflows with review and approval steps, integration with existing systems
Selected work
My work spans insurance, healthcare and financial markets, including multi-year implementation engagements, advisory work and executive education. Organizations I have worked with include RxSense / SingleCare, iptiQ (Swiss Re's digital insurance business), and Loeb Enterprises and its portfolio companies.
-
Contract intelligence in production
Built core components of a contract-intelligence system that reached production and cleared internal governance review.
-
Forecasting and customer-lifecycle systems
Forecasting, retention, customer-lifecycle and campaign-analytics systems for pharmacy-technology and healthcare workflows.
-
Customer analytics and data-science capability
Advised portfolio-company teams on machine learning and statistical modeling in healthcare and consumer analytics, and on building in-house data-science capability.
-
Live trading infrastructure
Designed, audited and operated research and execution infrastructure across equities and digital assets, including live systems for a regulated European digital-asset investment firm.
-
Investment research platform
At Infusive Intelligence, built a New York data-science office for investment research: quantamental workflows, forecasting, market-event models and a research platform for a public-equities strategy.
-
Executive AI education
Designed and delivered executive programs on AI, data science and analytics for senior leaders, including a large insurer, with selected participant projects moving into implementation.
Public record
- Book
Machine Learning for Trading, now in its third edition, with its open-source code; more than 100,000 professionals and learners taught.
- Open source
PDF-Bench compares document-extraction results; PDFsmith gives alternative parsers one interface; mdtoken and the Coding Agent Toolkit support AI-assisted software work.
- Recorded demonstrations
Two agent workflows run on public documents and synthetic records, recorded step by step to show the method: a supplier NDA review and a failed-replacement investigation.
How an engagement runs
I listen first, to understand your business, constraints and what success looks like. Then I deliver working systems early and improve them with real feedback.
-
1
Discover the highest-value constraint
Map workflows and bottlenecks, estimate the return from unit economics and total cost, and check feasibility, data readiness and compliance.
-
2
Design for trust and scale
Decide what belongs in deterministic code and what in a model or agent, plan data governance and security to your controls, and set cost, latency and approval requirements.
-
3
Deliver production systems
Build the application, integrate it with authentication, APIs and core systems, and ship the smallest useful release with cost, latency and quality baselines.
-
4
Operate with monitoring and control
Monitor cost, drift, safety and service levels, evaluate outputs automatically, and control changes to prompts, tools and models.
-
5
Improve for measurable value
Test against holdouts, analyze errors and user feedback, report return and risk to leadership, and support adoption.
How I work
-
Right tool for the job
Problems first; technology follows.
I start from the business problem and recommend the simplest solution that works. Sometimes the answer isn't AI, and I will say so.
-
Transparency and explainability
Every decision traceable, every outcome explainable.
Deterministic logic where possible, clear bounds on model uncertainty and decision traces stakeholders can audit, measured by business impact as well as model metrics.
-
Domain experts in charge
Your people shape the system from design through operation.
Subject experts define requirements, validate outputs and set success criteria. The system supports their reasoning, with review, approval and escalation built in.
Building your team's capability
Briefings for executives, workshops for leaders and pairing for engineers, so your organization can run and extend what we build.
- Executive AI briefings 2–3 hours
- Risk, return, operating model, vendor landscape
- Leadership workshops ½–2 days
- Use-case portfolio, prioritization, measurement plans
- Technical deep dives and pairing
- Retrieval and agents, evaluation, secure architectures
- Capability build-out
- Team setup, hiring and interview rubrics, operating playbooks
Who you work with
Stefan Jansen, Principal
I work on the strategy, architecture, evaluation and code. I came to machine learning through quantitative economics, financial markets and the business side of technology, then added the engineering depth to build the systems myself.
That sequence shapes the practice: first establish which decision or workflow should improve, then choose and build the approach that can be evaluated, governed and operated. I hold a Harvard master's in economics and public policy, a Georgia Tech MS in computer science and the CFA charter.