We engineer custom automation pipelines, ML models, and data infrastructure that replace manual labor with code. If a human is doing it repeatedly, we can probably automate it for less.
Every system we ship either saves you money or makes you money. We scope the ROI before we write the first line of code.
Multi-stage data pipelines that ingest public records, apply configurable filter chains, cross-reference obituary and property databases, and output scored, skip-traced leads with full contact data.
Custom-built sites engineered for conversion and search ranking. No WordPress templates. Hand-coded HTML/CSS/JS with structured data markup, semantic HTML5, and Core Web Vitals optimization.
Production and R&D machine learning systems built on PyTorch, TensorFlow, and Roboflow. From data labeling through model training to deployment. Real-time inference on video and image streams.
Programmatic outreach engines that handle email, SMS, and voicemail campaigns at scale. Template rotation, follow-up sequencing, response tracking, and CRM integration.
NLP-powered agent that handles inbound calls and texts, books into your calendar, sends confirmations and reminders. Replaces the receptionist role entirely.
Conversational AI on your website that qualifies visitors on budget, timeline, and intent. Routes scored leads to sales in real time. Runs 24/7 without breaks.
Automated invoicing with escalating follow-up sequences for late payments. Integrates with QuickBooks, Stripe, or your existing accounting stack.
Post-service SMS trigger that solicits Google reviews, follows up on non-responses, and intercepts negative feedback before it goes public.
OCR + NLP extraction from PDFs, scanned forms, emails, and screenshots. Outputs structured data to your CRM, ERP, or spreadsheet. Zero manual data entry.
Classifies inbound email by intent, drafts context-aware responses for routine queries, and escalates edge cases. Cuts email handling time by 70-80%.
Template-driven content generation, cross-platform scheduling, and engagement analytics. Consistent brand presence without the agency retainer.
Pulls recent sales, calculates ARV, generates a one-page comp report per property. What takes an analyst 30 minutes runs in seconds.
Scrapes county records, utility data, and satellite imagery to flag properties that go vacant. Automated alerts for investors and code enforcement.
We audit your current process, identify the bottleneck, and calculate the ROI of automating it before anything else.
System design, data flow mapping, API contracts. You approve the spec before we write code. No scope surprises.
Build, test against real data, iterate. You see working demos throughout. We ship when it's production-ready, not before.
Production deploy with monitoring. Monthly maintenance and optimization. The system improves over time, not degrades.
Production code. Real data. Measurable results.
Solo-built SaaS that gives baseball pitchers MLB-level scouting analysis from a CSV upload. Computes proprietary Stuff+, Location+, and Pitching+ grades using z-score normalization against 632 MLB pitcher profiles. Matches pitcher comps from 2,586 pitch profiles, optimizes pitch mix with platoon adjustments, and generates branded PDF scouting reports with Claude AI narrative analysis. Full monetization with 5 Stripe tiers ($0–$350/mo), post-checkout upsells, and a gated 11-step wizard UI.
Multi-format data ingestion auto-detects Trackman, Rapsodo, Statcast, and R data files across 50+ column name variants. Vision AI fallback parses scanned PDFs and phone screenshots via Claude Sonnet multimodal. Includes red flag safety screening for young athletes, 8-tab personalized training workbooks, and N8N webhook integrations for CRM automation.
6-stage Python pipeline: ingests county tax exports, applies multi-criteria filter chain (estate ownership, $10K+ delinquent, $100K+ market value), scrapes hctax.net payment history with Selenium anti-detection, verifies deceased owners via obituary NLP, and delivers scored leads on a weekly cron schedule.
Automated heir discovery across Tracers, Spokeo, and public records using undetected-chromedriver to bypass Akamai bot detection. OCR extraction from screenshot-based services via Tesseract. Outputs deduped, scored contacts with phone, email, and address. 86% hit rate across 268 subjects.
Computer vision system that analyzes live baseball video feeds to detect pitch clock start/stop conditions using object detection models trained on custom-annotated datasets. Built on PyTorch and Roboflow for data labeling, with TensorFlow for model experimentation. Real-time frame-by-frame inference pipeline.
Scraping infrastructure that discovers real estate businesses via DuckDuckGo Search API, visits each site + /contact, /about, /team pages, and extracts phone numbers, email addresses, and owner names using regex pattern matching and NLP heuristics. Processes 450+ businesses with incremental checkpointing.
Single-file static site with SEO-first architecture: JSON-LD structured data, semantic HTML5, Open Graph tags, and Core Web Vitals optimization. Zero framework dependencies. Hand-coded CSS with responsive breakpoints. Sub-second TTFB on static hosting.
Windows Task Scheduler orchestration that executes the full 6-stage pipeline, deduplicates against historical runs to surface only net-new leads, and delivers a delta report via SMTP. Includes error handling with screenshot capture on failure and automatic retry logic.
County tax records are public. Obituaries are public. Skip tracing APIs cost pennies per lookup. Here's the architecture behind a self-hosted lead gen system that replaces $2K/month data subscriptions.
A line-by-line cost breakdown of automating appointment booking, invoice follow-up, and review generation. With ROI calculations for businesses doing 50-500 transactions/month.
Technical deep-dive on web scraping infrastructure: anti-detection patterns, session management, human-like interaction timing, and the legal framework for scraping public records.
From dataset annotation to model training to real-time inference. Lessons learned building a computer vision system that processes live video at 30fps for temporal event detection.
How deceased-owner properties with unpaid taxes become the highest-margin deals in real estate investing, and why automated skip tracing gives you a 6-month head start on the competition.
Structured data, semantic HTML, Core Web Vitals, and why hand-coded static sites outrank WordPress templates in every measurable way. With real Search Console data.
Clarke Home AI is the technology and automation arm of Clarke Home Services Consulting. We don't resell SaaS tools or white-label other people's software. We write code. Custom Python, custom models, custom infrastructure — built around how your business actually operates.
Our stack runs from low-level web scraping (Selenium, undetected-chromedriver, anti-bot evasion) through data engineering (Pandas, NLP, OCR) to machine learning (PyTorch, TensorFlow, Roboflow) and production deployment (cron scheduling, SMTP automation, API integration).
We work with real estate investors, service businesses, law firms, title companies, and anyone who's paying humans to do work that code can handle faster, cheaper, and 24/7.
