Senior Data Scientist / Agentic AI & ML Systems / Co-Founder
I build production ML systems and agentic AI systems that combine LLMs, tools, context, and evaluation—from fraud detection and anomaly monitoring to MCP integrations and enterprise automation.
Senior Data Scientist on Walmart's global fraud prevention team, focused on production ML systems, agentic AI, evaluation, and measurable product trade-offs.
My background includes a PhD in Business Economics, with research on machine learning for time-series and sensor data. I also co-founded Naos Optics, an eyewear company and scaled it to EUR300k ARR. The common thread is turning real-world problems into practical systems people actually use.
MLOps
Co-built an end-to-end MLOps pipeline supporting 12 fraud models, each trained on roughly 100M rows, across billions of dollars in e-commerce transactions.
Built reusable Python packages and MLflow workflows that reduced model-development cycles from months to days.
AI enablement
Co-built a specialized agent skill that turns production ML workflows and engineering standards into context-aware guidance for data scientists.
Supports training, evaluation, model registration, and promotion.
Agentic AI / Harness engineering
Building practical LLM systems through tool integration, context, evaluation, and safeguards for agents that operate in real workflows.
Generates validated dashboards from BigQuery schemas and analytics intent.
Structures goals and working notes into tasks, owners, timelines, and updates.
Investigates logs and dashboards to surface anomalies and summarize evidence.
Turns briefs and source material into interactive presentation web apps rather than static slide decks.
Fraud
Fraud detection systems for evolving abuse patterns across customer and emerging agent-mediated activity, balancing loss prevention, false positives, operational cost, and customer experience.
Anomaly detection
Agentic anomaly detection for monitoring distribution shifts, combining time-series methods, LLM workflows, evaluation, and automated triage.
Commerce automation
Built a Shopify app from scratch for EU-compliant invoice generation, integrating with Billit and automating billing workflows.
Open source / MCP
Built a typed Python MCP server for invoice retrieval, exact payment-reference lookup, invoice creation, and payment updates in Billit.
Designed for safe agent use with environment-only credentials, idempotent writes, production safeguards, normalized outputs, and a fully tested API layer.
View on GitHubStartup
Co-founded an eyewear company scaled to EUR300k ARR, with product strategy, DTC growth, operations automation, and EUR150k in angel funding.
Visit siteOpen source / AI developer tooling
Three merged contributions across Code Puppy and its core plugins, spanning extensible model metadata, model-selection UX, cross-platform agent-completion notifications, and reliability testing.
View contributionsFraud prevention work across production ML, agentic anomaly detection, and analytics automation.
Product, brand, operations, automation, and international growth for Naos Optics, an eyewear company.
Causal ML and Bayesian modeling for European marketing and video advertising insights.
Machine learning for time-series, sensor data, smart agriculture, and self-supervised anomaly detection.
Client-facing analytics, computer vision, text mining, product roadmaps, and innovation workshops.
My research focused on machine learning for time-series and sensor data, with applications in anomaly detection, forecasting, self-supervised learning, and smart agriculture.
Google Scholar
At Naos Optics, my work spans product strategy, innovation, operations, automation, distribution, and growth. The company reached EUR300k ARR and raised EUR150k in angel funding while running a lean international DTC operation.
That work shaped how I think about building: useful products need user empathy, clear positioning, operational discipline, and systems that make the business easier to run.
Forecasting, anomaly detection, fraud detection, causal inference, foundation models, evaluation.
Python, SQL, JavaScript, PyTorch, AWS, GCP, BigQuery, CI/CD, Looker, APIs.
LLM applications, agent loops, skills, plugins, MCP, tool use, context engineering, multi-agent systems, evaluation, and safeguards.
Product development, user research, stakeholder alignment, entrepreneurship, go-to-market, international growth.
Open to conversations about AI product work, data science & ML roles, applied AI systems, startups, and technical collaboration.