Java Desktop Developer (Freelance short term project)

Job description

Company Description

Invito Software Solutions is a leading software development company providing world-class web and mobile solutions efficiently and cost-effectively.

We specialize in next-generation design patterns, responsive coding techniques, and rigorous quality assurance, resulting in high-quality apps with a high return on investment.

Our scalable services and customized engagement models cater to businesses of all sizes, from innovative startups to well-established companies.

We develop powerful and effective solutions that meet our clients' specific needs.

Project: Spring-Powered Desktop App for Invoice OCR → Clean CSVWhat you’ll build

• Desktop Java application (Java 17+) backed by Spring Boot
• Runs locally (offline-first), ships as a single native installer or fat JAR.
• UI options (pick one):
• JavaFX UI using Spring for DI (FXML controllers wired via Spring).
• Swing UI with Spring-managed services.
• Embedded local web UI: Spring Boot serves a UI (e.g., Vaadin or lightweight HTML/TS), shown in an embedded browser (WebView/JxBrowser).
• Invoice ingestion: drag-and-drop folders/files; PDF, PNG/JPG; multipage; batch mode.
• AI/OCR pipeline (pluggable via Spring beans):
• Local OCR (Tesseract) + layout/zone analysis, or
• Cloud OCR (AWS Textract, Google Vision) with retry/backoff, or
• LLM-assisted parsing to a JSON schema with guardrails.
• Field extraction (headers + line items): vendor, invoice #, dates, currency, taxes, subtotals/totals, PO, line descriptions, qty, unit price, amounts.
• Validation & review UI: show the document preview, highlight extracted zones, flag low-confidence fields, quick edits, autocomplete.
• CSV export: stable schema; normalize number/date/locale; export per file or batch.
• Rules & heuristics: vendor templates, regex fallbacks, learned patterns, per-vendor overrides.
• Quality metrics: confidence per field, accuracy dashboards, reject reasons, simple analytics.
• Offline by default, with optional cloud connectors for OCR/LLM and template sync.

Architecture (Spring-centric)

• App launcher: desktop entry point that bootstraps SpringApplication.
• Core modules (Spring beans):
• IngestionService: drag-and-drop, PDF/image decoding, page splitting.
• OcrService (strategy): TesseractOcrService, TextractOcrService, GcvOcrService.
• ParsingService: layout analysis, key-value detection, tables; optional LlmParsingService.
• TemplateService: vendor profiles, regex rules, learned mappings; local cache + optional remote sync.
• ValidationService: confidence scoring, anomaly detection, suggestions.
• ExportService: CSV writer (stable schema, locale normalization).
• MetricsService: capture confidence, errors, durations; local storage (SQLite/H2).
• DocumentPreviewService: render + zone overlays (PDFBox + image layers).
• Persistence: Spring Data (SQLite/H2 on disk) for runs, templates, audits.
• Config: Spring profiles (offline, cloud), YAML config for OCR provider keys, thresholds, CSV schema.
• UI layer:
• If JavaFX: Controllers get beans via Spring (custom SpringFXMLLoader), reactive updates via ApplicationEventPublisher.
• If Swing: @Component panels wired to services, event bus for updates.
• If Vaadin (embedded web): served by Spring Boot; package app with an embedded browser window.

Key user flows

• Drop files/folders → Ingestion queue (progress bar, cancel/retry).
• OCR + Parsing → Field map + line items + confidence per field.
• Review screen → Document preview with highlight boxes, editable fields, low-confidence badges, autocomplete from templates.
• Approve/Reject → Approved go to export queue; rejected capture reasons.
• Export → Single CSV or per-invoice CSV; schema versioning; logs.
• Analytics → Success rate, average confidence, common reject reasons, vendor leaderboard.

CSV schema (stable, versioned)

• schemaVersion, fileId, vendorName, invoiceNumber, invoiceDateISO, dueDateISO, currency, subtotal, tax, total, poNumber, …
• Line items (denormalized or separate CSV): lineIndex, description, qty, unitPrice, amount.
• Locale-safe formatting (ISO dates, dot decimal).

Offline/Cloud strategy

• Offline: Tesseract + local templates; everything runs without internet.
• Cloud (optional): switch OCR/LLM beans via Spring profile/env; graceful fallback to offline if unavailable.

Packaging & ops

• Distribution: jpackage/native installers (Win/MSI, macOS/DMG, Linux/DEB/RPM) or fat JAR.
• Logging: Spring Boot logging; per-invoice audit trail; export logs.
• Updates: optional auto-update check (profile-gated).

Security & privacy

• Local processing by default; redact PII in logs; encrypted at-rest store (configurable).
• For cloud calls: minimal payloads, signed requests, regional endpoints.

Nice-to-haves

• Hot keys and batch review UX.
• Template “learn” button: convert manual fixes into a saved vendor rule.
• Import/export of templates.
• Headless CLI mode: --input dir --output csv.

Deliverables

• Source code (Spring Boot project + UI layer).
• Packaged desktop installer(s).
• Sample vendor templates & test invoices.
• README + setup, profiles, and OCR provider docs.
• Test suite (unit + a few end-to-end fixtures).
• Short user guide (drop → review → export).

Qualifications

• Strong proficiency in Java and Spring Boot (DI, Spring Data, configuration, profiles).
• Experience building desktop UIs in JavaFX, Swing, or Vaadin (served by Spring) with responsive, user-friendly design.
• Skilled in troubleshooting, debugging, and performance tuning in Spring/Java.
• Familiar with OCR/AI integrations (Tesseract, Textract, Vision, OpenAI/Vertex) and robust parsing.
• Version control with Git; excellent communication; ability to work independently/remote.
• Bachelor’s in CS/Engineering (or equivalent experience).
• Freelance/contract experience is a plus.

Required Skill Profession

Computer Occupations