• Design and operate agentic automation workflows, building tooling that lets AI agents execute multi-step business processes reliably.

• Delivered full-stack web projects for 5+ independent clients — landing pages, e-commerce builds, and portfolio sites — owning requirements, timelines, and client revisions end to end.
• Integrated third-party REST APIs and payment services; deployed to Vercel and Cloudflare Pages; adopted React and Node.js on projects requiring dynamic functionality.

• Shipped front-end features for client web projects in a small team, consistently meeting production deadlines.
• Resolved cross-browser rendering bugs across Chrome, Firefox, and Safari via DevTools inspection; contributed to production delivery through pull requests, code reviews, and Agile sprint tracking.

• Designed and built a cyberphysical threat detection backend across transport, energy, water, and communications infrastructure using Isolation Forest — custom correction logic reduced DOS-event false positives across 4 infrastructure domains.
• Surfaced detections through a Flask REST API; mapped findings to MITRE ATT&CK categories; visualised live anomaly data on a real-time JavaScript dashboard. 68 TDD tests.

• Built a full-stack AI desktop app: FastAPI backend exposes a clean REST API consumed by a PyQt6 desktop frontend — upload PDFs, auto-generate flashcards and multiple-choice quizzes, and get step-by-step AI tutor explanations via Claude; analytics dashboard tracks score trends and weak topics.
• Wired a three-engine extraction pipeline — PyMuPDF, pdfplumber, and Tesseract OCR — handling digital and scanned PDFs; SQLAlchemy + SQLite persistence, Albanian/English i18n toggle, and a full backend and API test suite.

• Built and deployed a production full-stack webhook service integrating WhatsApp Business with an Odoo inventory backend over XML-RPC — Node.js + Express backend with PostgreSQL persistence, requests authenticated via HMAC-SHA256, rate-limited per IP, and intent-parsed by Claude AI for fuzzy product search.
• Maintained 95%+ test coverage across unit, integration, and end-to-end webhook message test suites; service handles live production traffic with full observability of the request lifecycle.

• Architected an 11-crate Cargo workspace: hybrid HTTP/headless rendering, CSS selector and XPath extraction engines, Tokio async worker pool, per-host token-bucket rate limiting — delivered as a single 9 MB self-contained binary with 33 tests across 7 crates.
• Shipped two parallel UI surfaces — a ratatui live TUI and an axum SSE web dashboard with embedded HTML; SQLite WAL-mode persistence guarantees crash-safe frontier recovery with zero scrape-data loss on unexpected exits.

• Compiled a Rust security service to WASM and deployed it to Cloudflare Workers — JWT validation (HS256/RS256) and sliding-window rate limiting at the network edge with zero cold-start; rate-limit state in Cloudflare KV cuts one full auth round-trip per protected request.

• FloraCare — Android plant care app; Kotlin + Jetpack Compose, Perenual species API, weather-aware care scheduling. (github.com/IsliBasha/FloraCare)
• publer-mcp — TypeScript MCP server, 13 tools, three-package pnpm monorepo, Zod-validated shared types, full mock mode. (github.com/IsliBasha/publer-mcp)

Software engineer with 3+ years of production experience across full-stack TypeScript/Node.js, Python, and systems programming in Rust. Built and shipped: a production webhook service handling live WhatsApp Business traffic (Node.js + Express + PostgreSQL, 95%+ test coverage); an AI study companion with PDF ingestion and LLM-powered tutoring (FastAPI + Claude); a cyberphysical anomaly detection system for smart-city infrastructure (Flask + REST API + TDD, 68 tests); and a Cloudflare edge security service compiled to WASM. CS graduate (Polis University, 2026) — comfortable owning features end-to-end, working across ambiguous requirements, and shipping clean, tested code fast.

• Thesis: ME-DT Framework — cyberphysical threat detection for smart-city infrastructure using machine learning anomaly detection