Ecommerce Price Intelligence & IT Command Center

Fragmented operations, no single view

The client is a B2B company selling specialized industrial products online. They had several problems running in parallel:

Competitor pricing was tracked manually. Someone would visit competitor websites, note prices in a spreadsheet, and compare against their own catalog. Inconsistent. Slow. Usually outdated by the time the spreadsheet was shared.

Server infrastructure was managed ad hoc. Multiple production web applications shared a single server. Different frameworks, different configurations, different deployment methods. No dashboard. No monitoring. If something went down, they found out when a customer called.

Cloud workspace security was unaudited. OAuth token grants, file sharing permissions, multi-factor enrollment, stale accounts. Nobody had reviewed any of it.

One platform, four modules

We built a centralized IT command center that consolidates all four concerns into a single internal application.

Competitor Price Monitoring. An automated scraper pipeline that crawls competitor websites, uses AI to extract structured product data from unstructured pages, and matches products against the client’s catalog with confidence scoring.

The extraction is not simple scraping. Competitor product pages have no consistent structure. The AI parses each page using few-shot prompting with domain-specific examples, extracting product identifiers, price, safety ratings, specifications, and attributes mapped to industry standards. A confidence scoring system flags low-certainty matches for human review instead of blindly accepting them.

Server Command Center. A dashboard for the production server hosting multiple applications. Service health monitoring across the web server, application runtimes, database, and containers. Disk usage tracking. Operational visibility that did not exist before.

Cloud Workspace Security Audit. Integrates with the workspace admin API via service account with domain-wide delegation. Audits DNS records, file sharing exposure, OAuth token grants, group membership, multi-factor enrollment, and stale accounts. Includes write actions: the admin can suspend users, revoke tokens, and remove admin roles directly from the dashboard.

Uptime Monitoring. Self-hosted uptime tracking for all the client’s web properties with alerting.

The technical build

The platform runs on a modern React framework with PostgreSQL (self-hosted with connection pooling) and a type-safe ORM. AI extraction uses a lightweight model optimized for structured data extraction at a fraction of the cost of larger models.

Authentication uses HMAC tokens with server-side enforcement. The entire application runs in Docker with a reverse proxy and automated SSL.

Development was orchestrated through an 8-agent AI team (PM, platform engineering, scraper engineering, AI/ML engineering, infrastructure, and QA), each with defined scope and review obligations.

Results

Competitor pricing intelligence on autopilot. No more manual spreadsheet tracking. New competitor prices are extracted, matched, and surfaced in a review queue automatically.

Complete infrastructure visibility. Multiple applications monitored from one dashboard. Server health, disk usage, and service status visible at a glance.

Cloud workspace hardened. First-ever security audit of the client’s workspace environment. Stale accounts identified. Excessive OAuth grants revoked. Multi-factor gaps closed.

Self-hosted, cost-controlled. Database migrated from a managed service to self-hosted Docker with connection pooling, reducing database costs while maintaining full operational ownership.

Engagement model: built and deployed. We built all four modules and deployed them to the client’s infrastructure. The client operates the platform independently. The system runs on their hardware, under their control.