Zero-downtime migrations, consolidation and performance rescue on multi-terabyte transactional systems — and the warehouses, data lakes and ETL/ELT pipelines they feed. Done in production, with the business running. Three decades of the depth most modern architects never built.
Every company has one: the database that runs the business and terrifies everyone. The migration postponed for years because nobody will own the risk. The system that collapses under every peak. The estate of servers accumulated through a decade of “temporary” decisions. The application whose authors left long ago, with the business logic buried in a thousand stored procedures.
The standard industry answer — “rewrite it” — is usually the most expensive and most dangerous option on the table. The engineering answer is to migrate, consolidate and optimize what runs the business, without stopping it, with verification instead of hope.
The discipline is always the same: empirical verification, rehearsed cutovers, performance proven under production load. And the hands are mine — I write the scripts, run the migrations and execute the cutovers personally. The accountability stays with me.
Version upgrades, platform moves and server consolidation executed hot, on live systems. Method: 100% test coverage of migrated code, full data and view verification, performance validated under production-equivalent stress, rehearsed cutover.
Track record includes CMMI-III environments and systems that absorb a whole day's demand in minutes-long traffic bursts. Zero incidents.
Multi-terabyte transactional databases: partitioning, indexing strategy, T-SQL surgery on business logic, aggregate architectures for systems ingesting hundreds of millions of events per year.
When the report takes hours and the hardware is fine, the problem is the code — and it's fixable.
Warehouse feeds and replicas that stay correct: AWS DMS orchestration, custom integration controllers, change tracking, checksum-verified incremental sync — automated to remove the human error from the loop.
I've published my own tooling here: sqlDataTrans, czAsyncSQLServer, Git version control for SQL Server.
Mission-critical code thousands of users depend on — elusive bugs, integration failures, data corruption, performance collapses. Including orphaned systems whose original developers are long gone: study the code, document it, then extend it safely.
Brought in when others couldn't crack it. That's usually the brief.
The analytical layer, built by someone who owns the transactional side of the pipe: star schemas, metadata-driven ETL/ELT frameworks written hands-on in Python, Spark/PySpark and T-SQL, incremental loads on change tracking, and lakehouse platforms — Microsoft Fabric, Snowflake, Databricks and Dataiku (full certification suite).
Pipelines age well when their author knows what the source database is doing at 2 a.m. For the AI systems on top of that data, see the AI & Data practice.
100% remote, in English or Spanish, for clients primarily in North America and Europe. The entry point is almost always the audit — small, fast, and it tells you the truth.
Short, fixed-scope review of the estate: real risks, performance headroom, consolidation opportunities and a defensible migration path — written for both engineering and leadership.
Full ownership of a zero-downtime migration or consolidation: plan, verification framework, rehearsals, hot cutover, and my name on the outcome.
For systems that are collapsing under load or growth: diagnose the real bottleneck in the code and the schema, fix it in production, leave the evidence.
Ongoing senior authority over a database estate or critical application — architecture decisions, code review, the person your team calls before touching production.
High-skill hands-on — architecting and writing the code myself · Advisory — senior counsel to your team and leadership · Part-time — a fraction of my week, sustained over time. Any length — from a one-off audit or a few weeks of hands-on code to the multi-year full-time contracts that have been my usual shape.
A selection from the full project history — each row links to the detailed record.
Platform depth you can verify: every Dataiku certification that exists, a speaker slot at Snowflake's flagship conference, and the Microsoft Fabric training plan I designed and taught to an entire Data & AI practice.
Core Designer (Apr 2024) · Advanced Designer (May 2024) · ML Practitioner (May 2024) · MLOps Practitioner (Jun 2024) · Developer (Jul 2024) · Generative AI Practitioner (Jul 2026). Every certification Dataiku issues — credential IDs verifiable on LinkedIn.
Represented Celestial Systems as a speaker and as its Principal Data & AI Architect at Snowflake Summit in San Francisco, 2025 — the platform's flagship conference. More talks and appearances →
As principal Data & AI Architect, designed and personally mentored the complete Microsoft Fabric training plan for Celestial Systems' Data & AI team — the practice that then delivered Fabric to production clients.
The questions that decide whether a migration gets approved — answered straight.
Yes, as an engineering discipline rather than a slogan: 100% test coverage of migrated code, full data and view verification, performance proven under production-equivalent load, rehearsed hot cutover. Delivered that way for Telepizza and ALK Abelló. Zero incidents, on record.
Deepest on SQL Server — multi-terabyte transactional systems, partitioning, T-SQL at business-logic scale. Cross-DB reach across Oracle, PostgreSQL, DB2, Snowflake and more (including years modernizing a database IDE that supports 30+ engines), plus NoSQL, vector databases and OpenSearch.
That's a specialty. I've repeatedly taken over undocumented production systems with no support from the original developers — studied the code, documented the platform, then extended it safely while thousands of users kept working.
A conversation through LinkedIn, then usually a short health audit of the estate — risks, performance headroom, migration path — before committing to the larger project. Remote, primarily North America and Europe.
Yes. This practice covers the whole path from transactional source to analytical layer: ETL/ELT design — written hands-on in Python, Spark/PySpark and T-SQL — change-tracking incremental loads, star schemas and lakehouse platforms — Microsoft Fabric, Snowflake, Databricks and the full Dataiku certification suite. When the goal is AI on top of that data, this connects directly with the AI & Data practice.
That's exactly the kind I've spent three decades with. Tell me about it — the audit that answers “how bad is it really?” takes weeks, not quarters.