Pourya Shahroudi

Gryphon Financial Group, Inc., Morgan Hill, CA

• Own the application at the center of the company's recovery business. Internal teams use it to process supplier invoice data, identify recoverable credits, and generate the invoices sent to Fortune 500 and government clients. It is an ASP.NET Core backend on .NET 8 with an Angular 17 front end, and I handle design, implementation, tests and release.
• First engineer at the company to use AI-assisted development, and took the proposal for it through management.
• Leading the migration of company source control from TFS/TFVC to Azure DevOps Git (in progress). Mapped dependencies across a legacy codebase nobody had a current picture of, wrote the migration plan, and am designing the branching and deployment process that goes with it, introducing pull requests to a development flow that has never had them.
• Audited a workspace holding 20 years of history across multiple projects, separated real source from build output and package folders, and defined a multi-repository layout that preserves dependencies between the main application and the legacy Identity solution.
• Built profile-based configuration covering Default, Development and QA environments, including environment-specific Excel templates, with automated tests that fail the build on a bad profile before it reaches QA.
• Folded the Angular build and publish steps into the .NET deployment process, so front end and back end ship as one artifact.
• Kept authentication working through the modernization by running the .NET 8 services against the legacy .NET Framework 4.6 Identity provider.

SonicWall, Milpitas, CA

• Led design and development of the QA dashboard used by 30+ managers and QA leads to track test coverage and process health across the organization, replacing hand-assembled reporting. Containerized and deployed it on Docker Swarm.
• Owned the public SonicWall Product Security Incident Response advisory site (psirt.global.sonicwall.com), covering design, development and ongoing publication of security advisories to customers.
• Built a Django admin layer where leads compose reports from filters and get the matching Power BI report on demand.
• Consolidated test and defect data from TestRail, ApTest and Jira into one reporting model, replacing lookups across three systems with a single view.
• Built a Grafana dashboard aggregating device error and warning messages from the test environment into a live view of network health and the errors behind it.

Fujitsu Network Communications, Inc., Sunnyvale, CA

• Developed the network management system that provisions and monitors Fujitsu optical network elements including transponders, ROADMs and muxponders, working across the Java backend and the JavaScript front end.
• Contributed GUI and server API changes to a performance project that let a single request load 5,000+ network services, a volume the previous implementation could not handle.
• Built GUI and backend support for the multilayer path computation engine, including signal health checks between source and destination using bit error rate measurement.

Intelerad Medical Systems, Montreal, Canada

• Built application-layer features on a distributed Linux-based medical imaging platform used by radiology practices, from requirements analysis and high-level design through spec, implementation and customer support.
• Added a document scanning workflow letting technologists scan, edit and upload documents to the server for later review in the image viewer.

Python, LLM agents, Google Search Console and Business Profile APIs, SQLite, GitHub Actions. labouchee.us, and the case study.

• Built and maintain the bakery's website and the more than 20 scheduled jobs behind it on a self-hosted Linux server, with unit test suites that run in CI on every change. The watchdog over all of it is plain Python with no LLM, since an agent can report success when the job under it failed.
• Wrote a monthly SEO agent that reads Search Console and Google Trends data and proposes up to five site changes, each required to cite the data behind it. Code-level changes arrive as a pull request for a person to review and merge, and a later job judges each changed page by its traffic before and after, with the rest of the site as the baseline.
• Built a weekly lead finder for corporate dessert orders. An LLM picks from a list of already vetted Bay Area companies and drafts a note for each, and code rejects any lead naming a company outside that list or citing evidence that is not that company's. Sending stays manual.
• An LLM drafts a reply to each new Google review and a weekly Business Profile post. Nothing is published until a person approves it.
• Track whether AI assistants recommend the bakery: a monthly job puts a fixed set of customer questions to Claude with web search, with ChatGPT through the OpenAI API on demand, and records whether the bakery is named and where.
• Built an order-intake pipeline that read the business mailbox, created a calendar event for each incoming order, and ran a nightly LLM agent classifying orders as confirmed, declined or cancelled from the email thread. Retired it after measuring the classifier against real outcomes, because customers also confirm by text and Instagram, which email never sees.

JavaScript, Chrome Manifest V3. Published on the Chrome Web Store.

• Built and published a Jalali calendar with Gregorian and Hijri date conversion and Iranian public holidays. All data ships inside the extension, so it makes no network requests and asks only for the storage and alarms permissions.