Kaltrep

Turning data into direction — and complexity into competitive advantage.

A Data Driven Investments Company

See What We Do
Scroll

What We Do

Where engineering depth
meets domain expertise.

We design and build data systems that actually ship — pipelines, Lakehouse architectures, analytics platforms, AI systems, and the governance frameworks that make them last. Most of our work is in oil and gas, where we bring 18+ years of operational engineering context alongside the technical depth.

01

Data Platform Engineering

I build the foundational infrastructure that makes analytics possible at scale — Lakehouse migrations, pipeline architecture, Unity Catalog governance, and CI/CD for data teams. Most clients come to me when their SQL warehouse is collapsing under its own weight or when a cloud migration has stalled.

Databricks Unity Catalog Azure Data Factory Lakehouse Architecture Pipeline Engineering
02

Analytics & Business Intelligence

From executive dashboards to self-service platforms, I design analytical environments that business users can actually operate. That includes BI architecture, Power BI and Spotfire builds, and the enablement programs that get adoption above 10%.

Power BI Spotfire Databricks Dashboards Self-Service Analytics Data Strategy
03

Artificial Intelligence

I build AI that does the work — not demos that gather dust after the meeting. Agentic systems that run workflows end to end, foundation models hosted and fine-tuned on your own data, and custom MCP servers that wire them into the tools your team already uses — Databricks, Salesforce, Box. From chat interfaces to fully automated pipelines, every build is measured the same way: hours returned and faster decisions.

Agentic AI LLM Fine-Tuning Model Hosting Custom MCPs Chat Interfaces AI Workflows
04

Oil & Gas Operations

My O&G background runs from the wellsite to the boardroom. Drilling Engineer. Completions Field Supervisor. Operations Planning Engineer. Then the architect of the digital systems that gave executive leadership real-time visibility into all of it. I've worked upstream from every angle, which means I scope problems correctly the first time — not after three discovery sessions.

Upstream O&G Enverus WellView SCADA GIS / Esri Capital Programs

About

Jeff Steele — Kaltrep

Jeff Steele

Former Director of Business Intelligence at EQT Corporation, where I led the company's full Databricks Lakehouse transformation — a 35-person analytics organization, 260K+ monthly integrations, and a citizen developer program that scaled to 170+ active participants. Before data, I was an O&G engineer and a Marine officer. I build systems that last because I've operated the ones that didn't.

Full profile →
18+ Years in data leadership
260K+ Monthly integrations managed
40% Pipeline cost reduction delivered

Our Approach

Operators,
not just advisors.

Kaltrep brings the precision of engineers and the perspective of leaders who have sat in the seat — across enterprise data transformations, modern Lakehouse architectures, and full accountability for analytics strategy in complex organizations.

Every engagement delivers practical outcomes — clear, actionable roadmaps your team can execute, not slide decks that gather dust.

Recent Work

Delivering outcomes,
not just recommendations.

Financial Services  ·  Insurance
01 / 04

Automating the Insurance Quote & Renewal Workflow

An insurance broker managing 120+ clients was spending significant time per engagement manually wrangling data from multiple sources, filling out individual forms for each of 8 insurance carriers, and sending them via separate emails — before a single quote could even be returned. Renewal notices were tracked and sent by hand. We rebuilt the entire workflow inside HubSpot.

The Problem

Each quote required a broker to aggregate client data from disparate sources into a spreadsheet, then manually transcribe that data into a unique form for every insurance provider — attached to individual emails sent one by one. Renewal reminders were manually managed, creating risk of missed notices. Client engagement at events was tracked informally with no structured follow-up.

The Solution

We implemented a full HubSpot CRM solution with custom event lead capture forms, automated post-event follow-up workflows, and a broker-facing quote intake form that simultaneously distributes to all 8 insurance providers via API integrations — eliminating redundant data entry entirely. A 30-day policy renewal automation was deployed to replace manual tracking, ensuring no client renewal goes unnoticed.

120+ Contacts Managed
8 Carriers Integrated
100% Quote Submission Automated
0 Manual Renewal Notices
Built with HubSpot CRM API Integration Workflow Automation Custom Forms Event Lead Capture

Capital Program Intelligence for a Multi-Plant Energy Build

A power-generation company executing a seven-plant capital build was tracking budget, commitments, work-in-progress, and actuals across systems that didn't talk to each other — ERP, project scheduling, work management, and engineering tools. Leadership had no single, trusted view of cost across the program. We designed a Databricks Lakehouse to serve as the single source of truth linking every dollar across all seven plants.

The Problem

Capital cost data lived in separate platforms — ERP for actuals, the CRM for purchase orders and contracts, scheduling tools for the WBS, and work-management and engineering systems for the rest. Numbers were re-keyed by hand between systems, reconciliations were slow, and executives couldn't see budget-versus-actual across the program without weeks of manual consolidation.

The Solution

We architected a Databricks Lakehouse as the central hub — every system publishes to it and subscribes from it, with no application-to-application integrations permitted. A unified cost fact table joins budget, WIP, and actuals; a financial semantic layer standardizes metric definitions; and governed pipelines feed executive dashboards, accrual automation, and earned-value metrics (CPI, EAC, forecast variance). The same model scales across all seven plants.

20–40% Cost Reduction Potential
7 Plants, One Model
1 Source of Truth
0 App-to-App Integrations
Built with Databricks Lakehouse Azure Data Factory Salesforce NetSuite ERP Earned Value Mgmt

PropScout AI — Real Estate Investment Intelligence

Real estate investors waste hours manually pulling property data, estimating rents, and running the same return calculations one listing at a time. We built PropScout AI — an AI-assisted platform that screens properties against institutional-grade investment criteria automatically and hands the investor a ranked shortlist with the analysis already done.

The Problem

Evaluating a rental property means gathering data from multiple sources, estimating rent, and computing cap rate, cash-on-cash, DSCR, and more — by hand, one property at a time. It's slow, inconsistent, and doesn't scale across markets, and scraping listing sites directly isn't a compliant option.

The Solution

We designed a full-stack app that pulls licensed property and rent data through MCP-connected APIs, computes the five core investment metrics against user-controlled assumptions, and layers in AI enrichment — neighborhood scoring, short-term-rental demand, and risk flags. Results are ranked and one-click ready: each property generates a pre-filled deep link into the investor's analysis platform. Built end to end in Claude Code.

140M+ Properties Covered
5 Investment Metrics
17 Tunable Assumptions
1-Click Deal Analysis
Built with Claude Code React / FastAPI RentCast + Apify MCPs Google Cloud Gemini

Geospatial Site Selection for Energy Development

Choosing where to build an energy project means weighing dozens of competing factors across the landscape — terrain, sun exposure, proximity to infrastructure, grid tie-in cost, and more. We built a GIS-based analysis system that scores every candidate parcel against a proprietary model and surfaces the optimal locations automatically, replacing weeks of manual map review.

The Problem

Site selection was a slow, manual process — analysts layering maps by hand, eyeballing slope and utility data, and estimating grid distances parcel by parcel. The work was subjective, hard to repeat, and didn't scale across a region. Comparing thousands of potential plots across multiple states on a consistent, defensible basis was effectively impossible.

The Solution

We built an automated siting engine in QGIS that ingests slope maps, utility and infrastructure layers, solar exposure, population density, and grid tie-in cost, then applies a proprietary weighted scoring model to rank every candidate parcel. The output is a ranked shortlist of optimal plots with the trade-offs made explicit — turning a manual, subjective exercise into a repeatable, data-driven one.

1000+ Locations Scanned
3 States Analyzed
12 Counties Covered
6+ Map Layers Weighted
Built with QGIS Geospatial Analysis Slope & Terrain Modeling Grid Tie-In Analysis Proprietary Scoring Model

Get in Touch

Let's talk.

Whether you're evaluating a data platform, planning an O&G digital transformation, or not sure where to start — we'd welcome the conversation.

Connect on LinkedIn Send an Email