I focus on building solutions that meet real user needs and fit naturally into existing business processes. With experience in data analysis and modeling, I prioritize practicality over complexity—creating tools people can actually use. I collaborate closely with clients to understand their workflows, communicate clearly, and develop systems that support both their goals and the way they work.
I treat AI as a force multiplier: it accelerates discovery, drafts solutions, and unlocks quicker iterations, but every line it suggests is reviewed and vetted before it ships, and the process stays flexible—choosing the right mix of automation and craft—while grounded in a responsible, professional ethic.
Invoice Mailer
Why it exists: Finance teams were manually hunting invoices/Statements of Account across shared folders, zipping them, and pasting boilerplate emails. The app turns that ritual into a guided workflow with guardrails: pick a billing period, validate matches, preview zips, and send (or safely dry-run) in one place.
UX shape: A Tkinter desktop app with tabs that mirror the real-world flow—Settings → Email Settings → Scan → Zip → Send & Logs—so users always know what’s next. Mode is loud and color-coded (red Active vs blue Test) to prevent accidental real sends.
Core pipeline:
Settings: Users point to invoice/SOA folders, the client list (Excel/CSV), and an optional zip output folder. SMTP is captured once, with TLS on by default.
Email setup: Subject/body templates, sender name, reporter emails, and the target month/year live here. Reporter emails receive run summaries.
Scan: We ingest the client list into SQLite, regex-match invoice/soa filenames, and parse invoice/SOA dates from PDFs with PyMuPDF plus optional OCR fallback. A change report highlights added/removed records so operators see what changed before sending.
Zip: Each head office gets a per-client zip with its invoices plus the matching SOA. The preview table is your preflight check.
Send: Builds EmailMessage objects, attaches the zips, and sends via SMTP. Test mode skips the network and produces a full “would send” activity log; Active mode sends and optionally mails a run report to auditors.
Engineering choices:
Local SQLite cache (backend/db/db.py): Fast, embeddable, and easy to rebuild. Every run snapshots the old DB, rebuilds from disk, reports diffs, and cleans up the backup to keep state honest.
File/date extraction (backend/config.py, backend/utility/extract_pdf_text.py): Regex-driven filename patterns keep parsing explicit. PDF regions are configurable percent boxes; OCR fallback covers scanned PDFs. Dates normalize to ISO so month/year filtering stays deterministic.
Packaging emails (backend/utility/email.py): Templated subjects/bodies with {month}/{year} tokens; zips attached per client; reporter emails receive the activity log. SMTP TLS and AUTH are enforced when available, and dry-run mode keeps testing network-free.
Secure settings (SecureConfig): Config is encrypted at rest. On Windows bundles we prefer DPAPI; elsewhere we fall back to a Fernet key stored alongside the encrypted config. Credentials aren’t left in plain text.
Resilience: Missing matches simply skip sending for that client; invalid paths raise early; threaded send keeps the UI responsive; progress bar + log pane make long runs observable.
Developer ergonomics: A local SMTP debug server script plus dry-run logging makes it safe to iterate without touching production mail. Config/regex defaults live in code but are easy to adjust.
Value to users:
Eliminates manual file hunting, zipping, and emailing: one run covers the whole billing period.
Reduces mistakes: enforced month/year selection, preview tables, and bright mode banners make it hard to send the wrong files or send for real by accident.
Auditable runs: change reports and reporter emails capture what was sent (or would have been sent) for compliance.
Faster onboarding: Desktop UI with saved settings means anyone on the team can execute the process without shell skills.