Data platform & AI engineering
Most data failures never throw an error. They load zero rows, duplicate a join, or drift a vector space. They do it quietly, for months, feeding the dashboards your executives read and the models your business has already started trusting. I find those, fix them, and build the platforms that stop doing it.
A join fan-out inflated a curated dataset 2.5×. No job failed. No alert fired. The duplicated rows fed a production model for eight months before anyone questioned the counts. This is the shape of most data incidents: not an outage, but a number that quietly stopped being true.
Services
Usually several arrive together. The platform needs modernizing, something on it is already broken, and there is an AI initiative depending on both.
The model is the easy part. The hard part is the supply chain feeding it: training sets that reflect reality, embeddings that refresh when the source changes, retrieval that returns the right context, and monitoring that tells you when it quietly stops working.
I build that layer end to end, covering medallion curation, training-data preparation, embedding generation and vector indexing, inference orchestration, and drift detection backed by an auditable prediction log.
Medallion architecture, Delta Lake, Unity Catalog migration, and incremental loads that actually stay incremental instead of quietly reverting to full reloads nobody costed.
Also the unglamorous half that decides whether the platform is affordable: cluster right-sizing, job-chain consolidation, and cutting multi-hour runtimes down to something you can schedule a business around.
Something has been wrong for months and nobody can say what. Silent zero-record loads. Race conditions in parallel loops. A type promotion that widened a column. A filter hardcoded to a year that has already passed.
I find it, prove it with the data, fix it under your change process, then build the validation that would have caught it in the first week.
Press either button. Same job, same empty source table, two different outcomes for the people who depend on it.
Spatial data has its own failure modes. Projections that silently disagree, geocoding that drifts, boundary files that stopped matching the addresses joined against them, and volumes that outgrow the single machine they were designed for.
I build enterprise geospatial pipelines covering spatial enrichment, geocoding, feature extraction, reconciliation across point-of-interest, address, boundary and imagery sources, and distributed processing when scale demands it.
How engagements work
Most teams can't specify this work up front, because the problem hasn't been diagnosed yet. Asking you to scope it before we look would just move the risk onto you. So we don't start there.
I go through the platform, covering pipelines, orchestration, data quality, lineage and cost, and come back with what's broken, what it's costing you, and what to do in what order. You own the document regardless of what happens next. If we never work together again, it still has to be worth what you paid.
The assessment names the work and prices it: the migration, the AI data layer, the reliability framework, whatever it turned up. Defined deliverables, defined end date, no open meter. You know what you're buying because we just spent two weeks establishing it.
For teams that need senior data engineering continuously but can't justify a full-time hire, or that need someone accountable for the platform between hires. Usually follows a build, because by then I already know the system.
Selected outcomes
Client names are withheld under confidentiality. Every figure below is from a production system, across federal health, defense, aerospace, energy, payer and geospatial environments.
Rebuilt a nightly full-load procedure as a true incremental MERGE with change detection and structured logging.
Federal health agency
Designed and executed an archive purge cutting roughly 96% of retained volume, averting filegroup saturation on a system days from failing its writes.
Federal health agency
Caught a join fan-out silently inflating a curated dataset that fed a production prediction model, then corrected the upstream logic.
Global aerospace manufacturer
Traced a model's quietly degrading accuracy to vector metadata written to ephemeral storage, forcing a full index re-fit on every single run and drifting the embedding space underneath it.
Global aerospace manufacturer
Built parameterized data-quality automation covering seven pipeline families, wired across 17 orchestration tasks in restricted environments.
Defense analytics platform
Root-caused a type promotion that had silently widened integer quantity fields across three production datasets, tracing it to placeholder casting in a shared transformation service.
Defense analytics platform
Consolidated legacy scheduled jobs into orchestrated workflow chains ahead of a platform-wide governance migration.
Defense analytics platform
Recovered a pharmacy-benefit data backlog by exposing hardcoded date filters, a parallel-execution race condition, and a recursive delete hazard producing silent zero-record loads.
Regional health plan
Connected LLM embeddings to a PostgreSQL vector store with HNSW indexing, enabling hybrid retrieval across work-management and equipment data for an operations AI assistant.
National energy operator
Delivered outbound interoperability moving encrypted, standards-compliant payer data to a national interoperability platform on a daily orchestrated schedule.
Regional health plan
Workflow efficiency gain from a full-stack geospatial application, recognized with an agency award.
Federal geospatial agency
Built enterprise pipelines reconciling point-of-interest, address, boundary and imagery datasets, with distributed processing for volumes past single-machine scale.
Federal geospatial agency
Credentials
Client logos aren't shown here, because the engagements are covered by confidentiality and several sit inside federal programs. What follows is mine to show.
Sectors delivered in: federal health · defense and intelligence · aerospace manufacturing · energy and nuclear operations · healthcare payer · geospatial
Who you'd be working with
Tech Leap LLC is my consulting practice. Ten-plus years building data platforms in environments where being wrong is expensive: veterans' health records, defense security cooperation, investigative and geospatial intelligence, aerospace quality, nuclear operations, and payer claims.
I work hands-on. You get the person who writes the PySpark and reads the execution plan, not an account manager with a delivery team behind a curtain. That's also the constraint. I take a small number of engagements at once, and I'll tell you early if yours isn't a fit.
Most of what I'm hired for now sits at the seam between data engineering and AI: teams with a model or an LLM initiative that's blocked, not on the model, but on the data underneath it.
A short description of the platform and what isn't behaving is enough to start. I'll tell you within a day whether it's something I can help with.