Rendo veloce ciò che è lento,
riparo ciò che è rotto,
e costruisco ciò che non esiste ancora.

Sono Salvatore Pirillo, ingegnere software full-stack da Cuneo. Sviluppo software che funziona davvero. Dal refactoring di sistemi legacy a progetti IoT, mi piace risolvere problemi complessi con codice pulito.

Inizia un Progetto Come Lavoro

Performance-First

Codice che gira veloce, non codice che sembra impressionante

Costruito per Durare

Architettura pulita che non dovrai riscrivere

Consegna Rapida

Dai requisiti alla produzione, senza ritardi infiniti

About meme

Sono Salvatore, ingegnere software da Cuneo. Mi piace costruire cose che funzionano bene. Ho ottimizzato sistemi che ora girano 20 volte più veloci e gestiscono migliaia di utenti senza problemi.

Negli ultimi anni ho lavorato su progetti molto diversi: sistemi di navigazione per barche, piattaforme di trading, app di domotica, riconoscimento targhe. Ogni progetto mi ha insegnato qualcosa di nuovo.

Non sono il tipo che butta buzzword a caso - preferisco mostrarti quello che ho costruito. Ho scoperto che il codice migliore è quello che non devi più toccare. Per questo prima di scrivere codice penso a struttura, architettura e manutenibilità.

2024 - Presente

Ingegnere Software Fullstack

Salvatore Pirillo

Remote

Sviluppo siti e app per startup e piccole aziende. Gestisco tutto, dai requisiti al deployment.

2023 - Presente

Ingegnere Software Fullstack & Consulente Tecnologico

Tecnel MC

Ibrido

Lavoro su progetti diversi: sistemi per barche, domotica, riconoscimento targhe, piattaforme di trading. Seguo i progetti dall'inizio alla fine.

2016 - Presente

Digital Marketing & Content Manager

Marcograssoart

Ibrido

Gestisco i social per un artista. Da zero a 106K follower su Instagram. Ho curato la parte digitale della mostra a Milano.

2022 - 2025

Laurea in Ingegneria Informatica

Universitas Mercatorum

Remote

Costruisco basi solide in informatica e ingegneria del software.

La tua visione. Il mio codice.
Risultati veri.Risultati veri.

Un processo snello pensato per portare il tuo progetto dall'idea alla realtà in modo efficiente.

Capire

Ascolto cosa ti serve. Qual è il problema? Cosa non funziona? Cosa va costruito? Niente gergo aziendale, solo domande chiare.

Analizzare

Studio il sistema, trovo i colli di bottiglia, capisco cosa causa davvero i problemi. A volte non serve riscrivere tutto.

Costruire

Scrivere codice che funziona. Testarlo. Renderlo veloce. Il codice migliore è quello che non devi più toccare.

Consegnare

Mettere in produzione, monitorare che funzioni correttamente, sistemare quello che serve. Il lavoro non è finito finché non gira liscio.

App Web Moderne

FRONTEND

Costruisco applicazioni web veloci e responsive con React e Next.js. Da dashboard real-time a piattaforme e-commerce complesse, creo interfacce che gestiscono migliaia di utenti senza rompersi.

React / Next.js

TypeScript

Redux Toolkit

Tailwind CSS

Start Project

Performance & Scala

BACKEND

Ottimizzo sistemi che erano lenti e li rendo veloci. Da piattaforme di trading che elaborano milioni di calcoli a domotica che gestisce 1000+ dispositivi. So come trovare i colli di bottiglia e sistemarli.

Node.js / Express

Python / Django

MongoDB / Redis

WebSocket / MQTT

Start Project

Sistemi Connessi

IoT

Costruisco sistemi IoT che funzionano davvero. Protocolli NMEA 2000 per barche, braccialetti smart per la sanità, sensori ambientali. Edge computing con Raspberry Pi ed ESP32. Tutto comunica in modo affidabile.

Raspberry Pi / ESP32

NMEA 2000 / Modbus

Bluetooth LE

Edge Computing

Start Project

Intelligenza Pratica

AI / ML

Computer vision che riconosce targhe con il 98.5% di precisione. Algoritmi per rilevare cadute negli anziani. ML comportamentale che impara routine e rileva anomalie. AI che risolve problemi veri, non buzzword.

YOLO / Computer Vision

NumPy / Pandas

TensorFlow

OCR / Pattern Recognition

Start Project

Infrastruttura

DEVOPS

Metto in piedi sistemi che girano 24/7 senza problemi. Container Docker, switch gestiti con VLAN, sicurezza di rete. Monitoraggio, auto-recovery e strategie di backup. Mi assicuro che le cose continuino a funzionare anche quando qualcosa si rompe.

Docker

Architettura di Rete

Monitoraggio Sistemi

Backup & Recovery

Start Project

Real ProjectsProjects

10 systems I built for real clients. From IoT and maritime safety to AI-powered recognition and performance optimization. Each solved a concrete problem.

IoT & Maritime

NautiSafe

Maritime Safety System

A maritime safety system for yachts and fishing boats that had to decide in real-time whether it was safe to navigate or not. The problem? You need to combine tons of data: what type of boat it is (CE safety category), how it's built (stability, hull dimensions), what the weather is like (wind, waves, forecast), and where you want to go. I designed an algorithm that takes all this data and calculates a risk index. The boat communicates via NMEA 2000 protocols (the maritime standard for sensors and instrumentation) through Bluetooth Low Energy - low consumption and works well in marine environment with salt and humidity. The system runs on ruggedized tablets certified for marine use. It must work even without internet (in the middle of the sea there's no signal), so I implemented offline-first architecture: nautical charts are downloaded beforehand and cached intelligently. When you return to port it syncs. Besides safety, it handles navigation with vector maps, overlays for routes and bathymetry (sea depth), and integrates with onboard entertainment systems. Result: system deployed on luxury yachts and commercial fishing boats, operational 24/7 even in rough seas and bad weather.

NMEA 2000Bluetooth LEOffline-firstIP67 Hardware

Risk calculation: boat + weather

NMEA 2000 sensors via BLE

IP67 ruggedized hardware

Performance Optimization

The One

Home Automation Re-architecture

A home automation system controlling lights, cameras, sensors, and access in a facility with hundreds of devices. It was slow, crashed often, and would start to collapse with more than 100 devices. The client wanted to expand to 1000+ devices but the system couldn't handle it. I analyzed it thoroughly: the main problem was how they transferred data - continuous HTTP polling where each device asked for updates every second, even when nothing changed. With hundreds of devices, this meant thousands of requests per second that brought the server to its knees. I re-architected the system piece by piece using WebSocket for real-time communication (devices receive notifications only when needed, they don't continuously ask), custom binary serialization to reduce traffic by 80%, and Redux Toolkit to manage state predictably. I developed custom parsers for proprietary IoT device protocols and implemented a health monitoring system with auto-recovery. The system now uses hybrid architecture: local processing with Redis for speed (if internet goes down, the house still works) + cloud sync with MongoDB for backup and remote access. Firebase Cloud Messaging handles push notifications to users' smartphones when something important happens. I also solved a critical user identity synchronization problem between local and cloud using MongoDB ObjectId instead of separate UUIDs, ensuring total consistency. Result: handles 1000+ devices with <50ms latency and 99.9% uptime.

WebSocketRedisMongoDBRedux Toolkit

1000+ devices, <50ms latency

Network load -80%

Uptime 99.9% (was ~90%)

Jupiter trading platform backtesting dashboard with real-time candlestick charts, NumPy performance metrics showing 4-hour to 10-minute optimization, parallel strategy testing visualization, risk management indicators including Sharpe ratio and drawdown analysis

Performance & Trading

Jupiter Platform

Trading Platform Optimization

A trading platform for testing strategies on historical data (backtesting). The problem? Algorithms written in NodeJS took 4+ hours to test a single strategy. With thousands of strategies to validate, it was impossible to iterate quickly. I analyzed the code: the main problem was nested for loops processing millions of CSV rows line by line. Each market tick was read, parsed, and processed individually - a performance disaster. I rewrote everything in Python using NumPy for vectorized operations (instead of loops, calculations on entire arrays), replaced CSV with Parquet columnar format (reads only the columns needed, not the entire file), and optimized critical sections with Cython (compiles Python to C for native speed). The system now integrates with IG.com for historical data and live trading, calculates risk management in real-time during tests (so you see immediately if a strategy would make you lose everything), and uses Monte Carlo simulations to validate robustness. Parallel processing: tests 10 different strategies simultaneously on different CPU cores. Database designed to support future brokers. All sensitive data (API credentials, account data) is encrypted per regulations. Result: backtesting that previously required 4+ hours now completes in less than 10 minutes. 95% time saved, thousands of strategies tested in one day instead of weeks.

PythonNumPyReal-timeAlgorithms

95% performance gain (4h → 10min)

Real-time market data

Complex trading algorithms

AI & Computer Vision

Alibi

AI Face Anonymization

A tool to automatically blur faces in videos and photos, developed for those who need to publish content while respecting privacy (GDPR). The problem? Manual blurring is incredibly slow - in a 10-minute video with dozens of people it takes hours, and you always risk missing someone. I built a system that uses AI (computer vision neural networks) to detect faces and automatically apply blur. Works on videos and photos. The interesting part is validation: not every face-shaped thing is a face - could be a poster, reflection, or mannequin. I implemented a scoring system that reduces false positives by ~80%. Typical use case: you filmed an interview with anonymous witness. The system detects all faces in the video and blurs them in a few minutes. Or a shop has theft footage to give to police - system blurs customers and employees, leaves only the thief identifiable. Relevant for journalism (anonymous sources), public surveillance (GDPR requires blurring faces), research with sensitive data.

CNNComputer VisionBatch ProcessingGDPR

Auto face detection + blur

False positives -80%

GDPR compliance

ALPR license plate recognition system showing live multi-camera feed with YOLO11 detection boxes highlighting Italian license plates, OCR text conversion with 98.5% accuracy metrics, real-time transit log, and edge computing latency indicators under 100ms

ML & Recognition

ALPR System

License Plate Recognition

A system that automatically reads license plates from cameras. Used by parking lots, restricted traffic zones, and security systems to know who enters and exits. The problem? Commercial cameras cost €5k-10k each and are closed systems. I developed a custom system that works with normal IP cameras (€200-300) and recognizes plates with 98.5% accuracy. The AI (YOLO11) is trained on 50k+ photos of Italian plates in every condition: night, rain, dirty, angled. The system reads the plate, converts it to text (OCR), and automatically corrects typical errors (0 vs O, 8 vs B). Flexible architecture: either central server processing recorded video, or mini-computer on the camera (edge computing) processing in real-time with <100ms latency. Handles multiple cameras simultaneously. Web dashboard to see transits, manage white/black lists (e.g., parking subscribers), and statistics. Use cases: parking lots with automatic barriers, restricted zone access control, condos with resident lists, warehouses tracking vans.

YOLO11OCREdge ComputingReal-time

98.5% accuracy

Works with cheap cameras

<100ms latency

Web & Mobile

Preventivando

Quote Management App

A mobile app for craftsmen and small businesses that make quotes (windows, electrical systems, photovoltaics). The problem? Making quotes on site takes time: you measure, take notes on paper, return to the office, enter everything on computer, calculate prices, send PDF to client. Days of waiting. I developed an app that does everything on site in 5 minutes: enter measurements, select materials from catalog, app automatically calculates quantities and prices, generates professional PDF that you send to client on the spot. Client receives quote while you're still there. For photovoltaics I implemented specific algorithms: enter available roof surface (e.g., 50 m²) and client's annual consumption (e.g., 3500 kWh/year), app automatically calculates how many panels needed, which inverter to use, and if worth adding battery storage. Considers roof orientation, shading, and tax incentives. Quote comes out optimized. Integrated team management: employees clock in/out from app (GPS verifies they're at job site), mark hours on each job, request vacation. Owner sees everything in real-time: who's working where, how much jobs cost, if someone's running late. Result: quotes that previously took 2-3 days now done in 10 minutes. Conversion rate (quotes → confirmed jobs) increased 40% because client receives immediate response instead of waiting days.

ReactNext.jsPDF GenerationReal-time

Quick quote generation

Team collaboration

PDF export & templates

EMS business management platform dashboard displaying real-time inventory DataTable with barcode scanner integration, order processing workflow, customer management interface, automated quote generation system, and performance metrics showing 5x faster loading with 60% code reduction

B2B & ERP

EMS

Integrated Business Management

I completely restructured an existing B2B web management platform developed in React, transforming a system with problematic architecture and incomplete functionality into a modern, scalable and performant solution. I performed complete refactoring eliminating massive code duplications and creating a dynamic pages system with reusable React components that drastically reduce necessary API calls. I completely redesigned the user interface with a modern and responsive design system. I rebuilt the backend architecture implementing advanced security standards and optimized routes for superior performance. The web platform manages customers, orders, warehouse, quotes and billing with automation for barcode labels and document generation.

Next.jsMongoDBAPI IntegrationB2B

Inventory & order management

ERP integrations

Custom B2B workflows

E-Commerce & Integration

EMS Commerce

B2B E-Commerce Platform

A client company in Burkina Faso (Apex Invest BF) that distributes solar and electrical materials to resellers and wholesalers had an EMS management system that worked well for managing articles, inventory, customers and internal orders, but completely lacked an online sales channel. B2B customers had to call or send emails to see catalog and prices. With hundreds of products and complex pricing system (sector markups, VAT, personalized discounts, 3 currencies), managing orders manually was slow and limited growth. The main problem wasn't just creating an e-commerce, but integrating it with the ERP without duplicating data. Articles, inventory, customers and price lists were already in the management system - rewriting them would have been a synchronization disaster. A system was needed that used the same data but with different pricing logic: final prices depend on sector markup (configurable), article VAT, discounts assigned to customer, and user role (consumer sees VAT included, reseller/wholesaler sees it excluded). I developed a custom B2B e-commerce in Next.js 16 with React Server Components for performance, fully integrated with the management system's Express backend. The frontend uses Better Auth for modern authentication with approval system: consumers are auto-approved, resellers and wholesalers require admin approval before seeing prices. Zustand manages state with persistence for cart and user preferences. To handle authentication between Next.js and Express backend, I created an API proxy (/api/proxy/[...path]) that intercepts all backend calls, extracts the HttpOnly better-auth.session_token cookie and forwards it to the Express server. This solves the problem that HttpOnly cookies aren't accessible from client-side JavaScript, ensuring security (session token never exposed) and automatic CORS handling. Result: production-ready B2B e-commerce managing 1000+ products with variants, response time <100ms for catalog thanks to multi-level caching (Redis + request-scoped + HTTP headers), 99.9% uptime, user approval system with 3 roles, persistent cart with multi-device synchronization and real-time price validation.

Next.jsE-CommerceMulti-tenantERP Integration

B2B e-commerce

ERP integration

Custom pricing & bulk orders

IoT & Healthcare

Arcle

IoT Healthcare with Behavioral ML

An IoT healthcare startup I co-founded as an entrepreneurial experiment. The problem? Elderly people fall and often nobody notices for precious minutes. Needed a system that automatically detected falls and immediately alerted staff. I designed and developed a complete system: smart wristbands with accelerometer and gyroscope that detect falls with 99.2% accuracy in less than 200ms. Cameras with facial recognition to monitor common areas without violating privacy (no recording, only alerts if someone is unwell). Environmental sensors for temperature, humidity, air quality. Edge computing architecture: Raspberry Pi 4 as central hub on each floor, ESP32 for environmental sensors, Raspberry Pi Pico in wearable wristbands. Everything communicates via MQTT (lightweight protocol for IoT). If internet fails, local system continues working and still alerts staff in facility. Django backend collecting and analyzing data: learns residents' routines and detects behavioral anomalies. Flutter app for staff smartphones: real-time notifications, event history, dashboard. System was completely developed and tested with real devices, validating edge computing architecture and fall detection algorithms.

IoTMachine LearningHealthcareRaspberry Pi

Fall detection with ML

Behavioral anomaly detection

Real-time caregiver alerts

Security & Infrastructure

Critical Infrastructure

Banks & Security Systems

I built security systems for banks and luxury residences. Not just software: I pulled cables, configured switches, installed Hikvision/Safire cameras and Ajax systems. The bank's access control worked with temporary QR codes - I saved it by rewriting the system that crashed after few hours due to memory leak and missing input validation. For a jewelry store I did people counting and heat mapping with retroactive reports. In luxury apartments: PTZ cameras on Ajax NVR, IP video intercoms with remote answer, each apartment with its own VLAN. Everything segregated for security.

Network SecurityVLANsMonitoringCompliance

Bank-grade security

24/7 monitoring

Regulatory compliance

CODE.

CREATE.

CONQUER.

Tre parole. Infinite possibilità. Ogni riga di codice è un passo verso qualcosa di straordinario.

Hai un progetto?
Parliamone.

Se hai un sistema lento da ottimizzare, un'idea da costruire o codice legacy da sistemare, contattami. Sono disponibile per progetti freelance.

Disponibile per lavori freelance

Tempo di risposta: < 24 ore

Rendo veloce ciò che è lento,riparo ciò che è rotto, e costruisco ciò che non esiste ancora.