Selected Work
AI engineering and agentic systems for production data platforms
Problems I solve
AI engineering & agentic systems — Coding agents wired to your data platform, CLIs, and real schemas: Text-to-SQL, tool/API contracts, evals, run manifests, sandboxed execution. Agents that ship to production, not chat demos.
Migration assurance — Agent-driven cross-platform validation for Databricks/Snowflake and cloud moves: schema, row-count, key-distribution, and date-range checks across dozens of tables — in a dashboard the team trusts.
Warehouse cost & performance — AI-agent workflows that profile the real bottleneck, benchmark across data scales and warehouse sizes, and prove row-level correctness before rollout. Faster and still right.
AI enablement & change — Opportunity discovery, stakeholder alignment, PoC-to-production, and capability building — with governance and ROI, not hype. Led a DS/analytics team and an ~11M SEK Vinnova/EU portfolio.
Time series & telemetry ML — Production ML for sensor and telemetry data — activity recognition, anomaly detection, forecasting — fast enough for edge/IoT (ROCKET family) at state-of-the-art accuracy.
How I work
Advisory — A few sessions to scope an AI-agent architecture, review evaluation strategy, or pressure-test a plan.
Embedded principal / interim lead — Weeks to months on your team; I own deliverables and ship agentic systems alongside your people.
Hands-on AI engineering — I build the thing and write the code — agents, guardrails, evals, tool contracts — not slide decks.
Evaluation-first delivery — A clear success criterion up front: I prove correctness before I optimize, and show evidence before I ship.
Selected examples
Text-to-SQL Against a Real ERP
Built natural-language querying against a wide production ERP with enforced retrieval, deterministic guardrails, value-aware evaluation (Wilson confidence bounds), and a streaming full-stack TypeScript/React app. Rebuilt as a fully synthetic open-source version preserving the real failure modes.
SQL Pipeline Optimization
Redesigned a critical daily full-recompute pipeline into a validated incremental one — 5-10x faster with 50%+ compute savings and row-level correctness proof. Multi-scale benchmarks and parallel AI-agent workflows killed an attractive but broken optimization before it reached production.
Cross-Platform Migration Validation
Built a production-grade validation system for a large Databricks-to-Snowflake migration: automated comparison across dozens of tables, historical result storage, and a live team dashboard — from zero to deployed in about a week using a coding agent.
Code Migration Methodology
Developed a practitioner’s framework for large code migrations and clean-room-style rewrites with coding agents. Covers evidence-preservation, behavioral oracles, shortcut audits, and agent guardrails — drawn from multiple 10k+ line migrations.
Time Series Classification for Production
Delivered a time series classification system for industrial telemetry with state-of-the-art accuracy and 10-100x faster inference than deep learning baselines. Designed for edge and IoT deployment where latency and power matter.
Engagement fit
You should reach out if:
- You are building AI agents against real data — Text-to-SQL, coding agents, RAG over schemas — and need someone who has shipped agentic systems with evaluation discipline, not just prompts.
- You are a data platform or engineering leader who wants AI-agent workflows for migration validation, pipeline optimization, or analytics automation on Snowflake/Databricks/SQL databases.
- You are driving AI adoption and need someone to run discovery, align stakeholders, and take agentic PoCs to production — not just advise.
- You are a technical sponsor or architect who needs a senior AI engineer to lead or unblock a workstream.
- You need an interim team lead who can own a roadmap, manage stakeholders, and still write code.
- You have a time series or telemetry problem that needs production ML, not a research paper.
I typically work with enterprise teams in manufacturing, logistics, energy, and industrial IoT — but the methods transfer.