Technology & delivered work

Data, analytics, and AI. Built, shipped, and running in production.

Over the past two years we have migrated enterprise data warehouses to Databricks, built retrieval systems over millions of pages, run a governed team of AI agents, and shipped web and mobile products that put data and AI to work.

1,700+
Tables in the enterprise lakehouse migration we work on
47+
Retailer data APIs in production ingestion
2,000+
Hours of hands-on delivery in under two years
14
Web applications built and shipped
5+
Mobile apps, native Android and cross-platform
Selected work

What we have delivered

A selection of data, analytics, and AI work in production, from enterprise migrations to retrieval systems and governed AI agents.

Data EngineeringThe largest water company in the US

Enterprise data warehouse migration to Databricks

A core engineer on an enterprise-wide Azure Synapse-to-Databricks migration spanning more than 1,700 tables across Finance, HR, Customer Direct, and Sales & Marketing. We migrated the datamarts behind revenue, labor, profitability, promotions, incentive compensation, and acquisitions reporting onto a Medallion lakehouse, without disrupting the reports the business runs on.

1,700+
Tables in the program
50+
Tables and procedures migrated
~20K
Lines of T-SQL converted
  • Led the team that set up the Unity Catalog governance framework: catalog and schema structure, naming conventions, and the standards every migration domain now follows
  • Converted T-SQL stored procedures to Databricks SQL on a Bronze/Silver/Gold Delta Lake architecture, deployed with Databricks Asset Bundles
  • Built a Python conversion framework on Lakebridge, with custom pre- and post-processing, that automated the translation of 25 stored procedures
  • Parallel validation compares row counts and MD5 checksums on every table, with dashboards, Excel and CSV reconciliation reports, and CI exit codes that stop a failed release
DatabricksUnity CatalogDelta LakeLakebridgeAsset BundlesSpark SQLAzure SynapseT-SQLPython
Data EngineeringThe largest water company in the US

Retail partner sales and inventory ingestion

Owned retailer data integration end to end, from access negotiation to production: 47 or more vendor APIs across a national mass-market retailer's data-sharing platform and a national home-improvement retailer's vendor portal, replacing manual portal exports with scheduled loads.

47+
Retailer APIs and feeds
2 yrs
Day-by-day backfill
10/s
Rate-limited requests
  • Ingestion notebooks for 40+ APIs on the mass-market retailer's platform, coordinated with the retailer's own data team
  • Seven vendor performance portal (VPP) and inventory feeds from the home-improvement retailer, with OAuth 2.0, request batching, and rate limiting that respects the API limits
  • Full two-year historical backfill followed by automated daily and weekly incremental loads
  • Analysis-ready fact and dimension tables for sales, inventory, and forecast reporting
PythonREST APIsOAuth 2.0pandasDatabricksJupyter
Data EngineeringMigration and production support at enterprise scale

AI-accelerated, triple-validated delivery

Integrated Claude Code and Genie Code with the Databricks environment to auto-generate validation coverage for whole domains and to carry migration and bug-fix work at several times the manual pace. No AI-produced change ships until it passes three rounds of validation against real data.

2–3 days
Per domain validated, vs. weeks by hand
1–3/day
Procedures migrated, vs. 2–4 days each
2–4/day
Production bugs fixed, vs. 1–2 days each
  • Auto-generated Lakebridge validation YAML for entire domains, delivering work other teams had spent weeks on by hand
  • Three named rounds on every change — coverage (all relevant tables and objects), correctness (KPIs and logic against actual data), and end-to-end QA
  • Read-only dry runs on full production data exposed silent date errors and duplicate-producing merges that review alone would have missed
  • Vendor best-practice guides converted into persistent agent instructions, so every AI-assisted task follows the standards by default
Claude CodeGenie CodeDatabricksLakebridgeYAMLPython
Analytics & BIThe largest water company in the US

Executive dashboards and conversational analytics

Self-initiated after the migration: proposed executive analytics to a VP, then built AI/BI dashboards and natural-language Genie Spaces to his directives, giving leadership one big-picture view of the operation and a weekly scorecard.

7
Executive subject areas
Weekly
Executive scorecard
Ask
Questions in plain language
  • Dashboards for trade promotion management, transportation (cost, service, and fleet), customer base, plant downtime, branch inventory, retail water demand, and HR workforce health
  • Genie Spaces let leaders ask questions of the data in plain language instead of waiting for a report
  • Built on the governed lakehouse, so the numbers match the ones operations already uses
Databricks AI/BIGenie SpacesDatabricks SQLUnity CatalogDelta Lake
Analytics & BIU.S. Department of Education data

IPEDS higher-education data

Extracted the federal IPEDS 2023–24 database, which covers every U.S. college and university, from its Microsoft Access distribution into open, analysis-ready tables for institutional research and reporting.

2023–24
IPEDS collection year
All
Survey tables exported
  • Automated extraction of every IPEDS survey table from Access to CSV
  • Loaded into SQL for enrollment, completions, and finance analysis
  • Repeatable process for each new annual release
PythonSQLAlchemyPostgreSQLJupyter
Analytics & BIGreen Olive Tech product

BIForge: Power BI generation and migration

A tool that generates complete Power BI projects (semantic models, measures, and report pages) from database definitions and KPI specifications, and migrates existing reports between platforms such as Synapse and Databricks.

PBIP
Native Power BI project output
0
AI in the migration path
  • Deterministic SQL translation for migrations, so results are repeatable and auditable
  • AI used only to turn plain-language KPI intent into model definitions
  • Produces TMDL semantic models, PBIR report layouts, and Power Query M code
PythonPower BI (PBIP/TMDL)sqlglotPydanticClaude API
AI & RAGArabic and English question answering

Bilingual retrieval-augmented knowledge assistant

A RAG system that answers questions in Arabic or English from a curated library of classical texts, citing its sources and refusing to answer beyond them.

16,314
Indexed passages
32B
Parameter self-hosted LLM
2
Languages
  • Semantic search with sentence-transformer embeddings and a Chroma vector store
  • Answers grounded strictly in retrieved sources, with confidence thresholds
  • Self-hosted Qwen 2.5 model on GPU, so no data leaves the environment
  • Answer quality measured with RAGAS evaluation; served through a FastAPI service
PythonLangChainChromaDBsentence-transformersOllamaFastAPIRAGAS
AI & RAGMillions of pages, exact citations

Large-scale Arabic corpus search engine

Extracted a 6,982-work classical Arabic library from legacy Access databases into PostgreSQL and built a retrieval engine that returns exact, citable passages in milliseconds.

6.4M
Pages processed
8.9B
Characters indexed
624s → ms
Query time
  • 5.9 million chunks stored in PostgreSQL with pgvector
  • Benchmarked multilingual embedding models (e5, BGE-M3) against BM25 lexical search
  • Shipped the approach the measurements supported, rather than the fashionable one
PythonPostgreSQLpgvectorBM25BGE-M3e5
Agentic SystemsGoverned AI agents that design, build, and review software

Eight-agent engineering system

A team of eight specialized AI agents (architect, independent reviewer, backend, frontend, and pipeline engineers, plus marketing strategy, copy, and design) that delivers software under explicit governance.

8
Specialized agents
2
Independent quality gates
T1–T3
Risk tiers
  • Spec, review, build, and verify workflow; no agent ever scores its own work
  • Risk tiers decide how much review a change gets; high-risk work needs an approved spec
  • Every decision and gate verdict is logged, and the system learns from retrospectives
  • Governance-aware MCP server for GitHub records who changed what, and why
ClaudeModel Context ProtocolPythonYAMLJSON SchemaGitHub
Agentic SystemsGreen Olive Tech product

CodePrizm: autonomous codebase analysis

AI agents read a Git repository and produce architecture documentation, security audits, test suites, and user guides, including scans of data platforms for PII, secrets, and access control.

4
Deliverables per scan
2
LLM backends: cloud or self-hosted
  • Works with the Claude API or a self-hosted model for sensitive code
  • Data-platform scanner covers SQL, Tableau, and Power BI assets
  • Exports to PDF and Word for audit and handover
PythonFastAPIClaude APIOllamaPydanticWeasyPrint
ApplicationsCollege Links, AI EdTech platform

Load-testing an AI platform to production readiness

Specified the platform in full before any code was written, then ran seven phases of stress testing and tuning against its API until it held up under heavy concurrent load on a single server.

180+
API endpoints
27% → 99.51%
Success at 1,000 simulated users
341/s
Requests sustained
  • Seven phases of Locust stress testing at 1,000 simulated concurrent users, each one measured before and after the change it motivated
  • Async I/O, connection pooling, caching, and index work drove the gains, on one server rather than more hardware
  • Every feature built from a specification reviewed through an agentic review framework before coding
FastAPIPostgreSQLElasticsearchLocustNext.jsExpo
ApplicationsGreen Olive Tech infrastructure

Self-hosted AI platform

A self-hosted AI lab, a GPU workstation and a Linux server, that runs open models, this website, and production web applications, so client work that cannot leave an environment does not have to.

32B
Parameter model served on our own GPU
0
Data leaving this environment
  • Production web apps and open models behind Nginx and Traefik reverse proxies
  • Hosts the bilingual RAG assistant and an automated pipeline that turns a manuscript into a ready-to-publish book, cover and barcode included
  • The same stack we deploy for clients who need AI inside their own network; our hosted products that call third-party models run separately from it
LinuxDockerNginxTraefikOllamaComfyUIPrometheusGrafana
Applications & pipelines

Delivered Products

Web apps, Android and cross-platform mobile apps, and bilingual production pipelines, each running in production. A selection of the fourteen web and five-plus mobile applications we have shipped.

AI web + mobile

InvExtracts

Invoice OCR and AI extraction into structured data, with QuickBooks and Xero export.

FastAPI · Celery · PostgreSQL + pgvector · GPT-4o-mini · React Native · Kubernetes

AI web + Android

College Links

Academic platform for flashcards, lecture analysis, and exam preparation.

FastAPI · PostgreSQL · Elasticsearch · Next.js PWA · Expo (Android)

AI web app

Claims Court Assistant

Guides people through small-claims cases in Texas, California, and Florida.

FastAPI · React 19 · PostgreSQL · Redis · OpenAI · Stripe

Android app

PassPhoto

Compliant passport and ID photos on your phone. Published on Google Play.

Kotlin · MVVM · Gradle

Android app

PrimeScanX

Document scanner with on-device OCR and PDF export.

Flutter · Google ML Kit · Riverpod · SQLite

AI pipeline

Audiobook Production Pipeline

Turns manuscripts into mastered, retail-ready audiobooks with voice cloning and QC reports.

Python · PyTorch · XTTS · FFmpeg

Production pipeline

Book Publishing Pipeline

Manuscript to print-ready PDF, EPUB, cover art, blurb, and ISBN barcode in English and Arabic, with source validation gates.

Python · ReportLab · Stable Diffusion · Arabic shaping & BiDi

Production pipeline

Calendar Publishing Pipeline

Print-ready bilingual English and Arabic monthly calendars for multiple cities, with calculated prayer times, Hijri dates, and moon phases.

Python · ReportLab · CMYK color · automated QA checks

Capabilities

What we can build for you

The platforms and practices behind that work, available for your next project.

Databricks Lakehouse Engineering

From legacy warehouse to a governed, AI-ready lakehouse, built end to end.

We design, migrate, and run production workloads on the Databricks Data Intelligence Platform. We cover the full lifecycle: assessing legacy platforms, converting code, validating data parity, governing with Unity Catalog, and handing over automated, CI/CD-deployed pipelines and AI-powered analytics.

Medallion Architecture

Bronze, Silver, and Gold layers that take raw ingested data through cleansed, conformed tables to business-ready aggregates, with clear contracts and lineage at each layer.

End-to-End Migration

Full migrations from SQL Server, Azure Synapse, and other platforms: assessment, schema and code conversion, data movement, cutover, and decommission planning.

Lakebridge Migration & Validation

Databricks Lakebridge to analyze legacy code, transpile SQL, and reconcile source and target data, plus our configuration-driven (YAML) validation framework that checks row counts, schema, and data values automatically.

Unity Catalog Governance

Centralized access control, catalog and schema design, data lineage, auditing, and secure data sharing across workspaces.

Stored Procedure Migration

T-SQL stored procedures converted to Databricks SQL and PySpark, then validated with side-by-side output comparison and validation dashboards.

AI-Powered Dashboards

Databricks AI/BI dashboards that business users can explore in natural language.

Genie Code & Genie Spaces

Genie Code, the Databricks AI agent, speeds up building pipelines, notebooks, and dashboards. Curated Genie spaces let business teams ask questions of trusted data in plain English.

Notebooks

Modular, parameterized, version-controlled notebooks for engineering, analysis, and data science.

Jobs & Scheduling

Multi-task workflows with dependencies, retries, alerting, and scheduled or event-driven triggers.

Spark / PySpark

Distributed processing tuned for performance: partitioning, joins, caching, and Delta Lake optimization.

Databricks SQL

SQL warehouses for high-performance analytics, reporting queries, and BI tool connectivity.

Classic & Serverless Compute

The right compute for each workload, with cluster policies tuned to balance performance and cost.

Asset Bundle Deployment

Infrastructure as code with YAML-configured Databricks Asset Bundles, promoted from dev to test to prod through CI/CD.

Technology stack

Tools we use every day

Data platforms

DatabricksAzure SynapseUnity CatalogdbtSparkPower BI

Databases

PostgreSQLpgvectorSQL ServerChromaDBRedisElasticsearchSQLite

AI & ML

ClaudeOpenAIOllamaLangChainsentence-transformersPyTorchTesseractML Kit

Languages

PythonSQLTypeScriptKotlinDartPowerShell

Applications

FastAPINext.jsReactFlutterReact NativeCelery

Delivery

DockerKubernetesGitHub ActionsNginxPrometheusGrafana

Have a data or AI project in mind?

Whether it is a migration to Databricks, a dashboard, a retrieval system, or AI agents connected to your tools, tell us what you are working on.