The prototype system consisted of a Slack application connected to a Python backend API:

• Slack interface for user interaction, responding in-channel with contextual answers
• RAG pipeline: document ingestion, embeddings generation, semantic search, and LLM response
• MongoDB vector database for document embeddings and similarity retrieval

The bot could summarize internal guides, retrieve code snippets, and answer architecture questions directly in Slack.

Proof of concept delivered. The prototype validated the technical feasibility and internal value of Slack-native AI assistants.

Adopted for expansion. After my delivery, the company’s internal dev team took over to evolve the system toward production.

Improved knowledge accessibility. Employees could find answers to code and documentation questions without leaving Slack.

• Google Computer Vision API to extract product information (OCR) from the images.

• Data sanitized and uploaded to AirTable, then easily catalogued (thanks to its own AI tools).

• Free AI tools to remove text and background from the images (and ofc web-optimize them).

• AirTable extension framework (a React web app bolierplate as baseline) to better control the UI styling and the layout.

OCR pipeline. The Google CV API was able to accurately extract all information from images.

AirTable AI tools. The AirTable AI assistant was great at cataloguing, deduplicating, and even extending the data into related tables (ie: unique brands, categories, etc.).

Free online AI. There are several free tools avalable online, to process images in all possible ways (u)sing more than one for the same purpose can bypass usage limits, keeping a free solution).

Go-Live in 48 hours. Besides the AI tools, the AirTable React extension speeds up the development process, allowing a quick and easy go-live of the product catalog.

A Node.js middleware handled the complete flow of product data between Salsify and Storyblok:

• Automatic synchronization of product data, attributes, and assets
• Scalable architecture able to handle multiple brand websites
• React web admin UI for the middleware configuration, with integration settings controls and logs.

The project also included the initial product data gathering and normalization for multiple global brands, preparing them for digital publishing consistency.

Unified content pipeline. Product data, descriptions, and images now flow seamlessly from Salsify into brand websites.

Scalable architecture. New brands or websites can be added to the middleware with minimal configuration changes.

Reduced manual effort. The marketing team no longer updates product information manually across multiple CMS instances.

Design-agnostic integration. The system supports future redesigns or new brand websites without touching the data flow.

The solution was a dynamic and modular integration layer: add new channels and data sources to the sync without rewriting the code.

• High volume of data was regularly synced between SAP and Salesforce
• Custom triggers code in Salesforce to handle the presentation layer
• Pricing logic was duplicated in CRM to ensure the correct pricing in quotes

I improved the performance and reliability of the integration layer release after release, until it could safely move high-volume daily data between the systems without falling behind.

Reliable data flow at scale. Even with heavy daily throughput, data stayed aligned between SAP and Salesforce.

Operational visibility. Sales teams saw the latest operational data in CRM; operations didn’t have to chase sales for context.

Incremental resilience. Performance and stability were improved iteratively — I didn’t guess, I hardened against real production load.

We delivered a formal Enterprise Architecture assessment focused on Marketing, but written for non-architect stakeholders.

• Clear baseline architecture: what tools, data, and processes exist today
• Target architecture: how automated lead capture, nurturing, segmentation, and campaign orchestration should work
• Gap analysis: what’s missing (people, platforms, integrations, data governance)
• Roadmap: which capabilities to build first, and which can wait

The result wasn’t “buy tool X.” It was: here’s how your org evolves from manual outbound to scalable, data-driven Marketing — and in which order.

Shared language with leadership. Sales, Marketing, and Exec stakeholders got a common view of where Marketing automation fits in the bigger company architecture, not just as a “MarTech project.”

Phased investment plan. The roadmap was sequenced into realistic milestones (process, data, tooling, governance), so the company could start moving immediately without committing to a risky full rewrite.

Strategic clarity. Instead of “automation everywhere,” we defined which capabilities actually matter for their recruitment business model — and which do not.

The project established an official, supervised data bridge between the company’s TPM system and Salesforce CRM:

• Promotion data synchronized so both systems reference the same data
• Processes and validation rules in CRM tuned so sales teams enter clean, TPM-ready data

Smoother promotion management. TPM activity could move intp CRM with less retyping and fewer gaps.

Clean handoff between teams. Sales and Marketing now talked to each other through systems — not screenshots and emails.

I introduced a dedicated middleware service that automated the sync of customers and quotes between CRM and ERP:

• Bi-directional sync of customer data (accounts / contacts)
• Quote synchronization: sales quotes created in CRM are automatically synced in Epicor ERP via API, with consistent item references and pricing logic
• Conflict resolution and deduping logic to prevent duplicate customers

The middleware orchestrated the sync of data between the two systems, and stored integration state in a document database for synchronization resilience.

Single source of truth for customers. CRM and ERP now reference the same account records instead of forks.

Faster quote-to-order flow. Quotes generated by sales in CRM can move directly into ERP without manual re-entry or reformatting.

Operational clarity. Both sales and operations see the same pricing, contacts, and commercial terms for a given hospital / care provider.

Lower error rate. No more version mismatch between “what was quoted” and “what production is building.”

The integration unified the company's new eCommerce stack around a custom middleware layer. Orders flow from online stores into the ERP, CRM and Shipping systems, while customer and product data syncs into the CRM and PIM.

Over time, this evolved into an automation platform for the business, including full order fulfilment, shipping & returns

Centralized data flow. Orders, inventory, customer data, and shipping status are synchronized and visible.

Automation everywhere. Manual steps in fulfilment, returns, and product management were replaced by reliable, repeatable jobs.

Lower friction for the business. Teams work in familiar tools (ERP, Salesforce) while integrations keep them in sync.