┌────────────────────────────────────────────────────────────────────┐
│ @HighMarck17 │
│ backend · systems · data · security · agent tooling │
├────────────┬───────────────────────────────────────────────────────┤
│ building │ services · APIs · automation · IoT · agent tooling │
│ context │ production data in regulated medtech │
│ location │ Italy │
│ contact │ highmarck@member.fsf.org │
└────────────┴───────────────────────────────────────────────────────┘
// mente algoritmica — security and privacy by design, enforcement over good intentions, and code that has to prove it works.
I build backend services and APIs, automation, IoT devices, and the tooling that lets coding agents work against real systems. By day I look after production data in a regulated medtech environment, where every record has to stand up to an audit; that discipline of traceability and verification shapes how I design everything else. Most of my development now runs through agentic tooling, which I use daily and build myself.
languages: [Python, Rust, C++, SQL, Bash, PowerShell, Java]
backend: [Django, Django REST Framework, OpenAPI, pytest]
data: [SQL Server, PostgreSQL, ETL, MES/ERP integration, Tableau, SAP BO WebI]
infra: [Linux, Docker Compose, Nginx, Gunicorn, GitHub Actions, iptables/NAT]
hardware: [Raspberry Pi, Arduino, ESP32, RFID/NFC, sensors]
security: [applied and post-quantum crypto in Rust, hardening, network auditing, CTF]
agentic: [MCP servers, Claude Code, OpenCode, subagents, hooks, skills]
local_llm: [Ollama, LM Studio, Qwen, DeepSeek]
practices: [TDD with mutation testing, ADRs, docs-as-code in CI, threat modelling]
exploring: [multi-agent orchestration, agent security]work/
├── backend/
│ ├── services, REST APIs with OpenAPI contracts, system integrations
│ ├── web applications on Django and Django REST Framework
│ ├── MCP servers: the integration layer between LLM agents and real systems
│ └── database design, SQL, testing with pytest
├── automation/
│ ├── Python, Bash and PowerShell scripting
│ ├── extraction, scraping and processing pipelines
│ └── GUI automation and scheduled desktop tasks
├── data/
│ ├── integration across heterogeneous systems (MES / ERP)
│ ├── master data, quality rules, reconciliation, traceability, ETL
│ └── reporting and dashboards for operational decisions
├── hardware/
│ ├── C++ on Arduino and ESP32, GPIO on Raspberry Pi
│ ├── IoT devices: RFID/NFC access control, door entry units,
│ │ encrypted LAN camera links, environmental monitoring
│ └── serial and network protocols on constrained devices
├── security/
│ ├── applied cryptography in Rust: Signal-style constructions,
│ │ post-quantum key exchange
│ ├── hardening public endpoints: rate limiting, honeypots,
│ │ brute-force protection on auth
│ ├── offensive practice: CTF web challenges, recon, injection testing,
│ │ findings written up with remediation
│ ├── anonymity: Tor transport and onion services
│ └── trust boundaries and permissions in agentic systems
└── systems/
├── Linux configuration, hardening and deployment
├── Docker Compose, Nginx and Gunicorn behind TLS, GitHub Actions
└── routing, iptables/NAT, Ethernet bridging across Linux and Windows hosts
Most of my work is not public: code written for an employer in a regulated environment is not mine to publish, and unaudited cryptographic implementations stay private on principle.
building ▸ home monitoring for fragile people on Raspberry Pi: on-device
computer vision, local AI as a verification layer, no video
recording, nothing leaves the house
▸ a post-quantum, peer-to-peer encrypted chat in Rust with no
central server, private until it has been audited
exploring ▸ multi-agent orchestration frameworks and the security of agents
that hold real tool access
open to ▸ consulting and collaboration on backend, data and security work
◆ harnesses
Claude Code and OpenCode as daily drivers: workflows, subagents,
path-scoped rules, skills, MCP servers as the integration layer between
agents and real systems
◆ persistent context
a layered memory system for coding agents: a small always-loaded core,
warm and cold layers on demand, a curator subagent for session briefing,
hash-based ratification with drift detection on the core
◆ enforcement over prose
memory files are context, not guarantees. Constraints that must always
hold live in PreToolUse hooks, permission denials and pre-commit checks
◆ model selection
local open-weight models where data must not leave the machine, frontier
models for the hard passes, cheaper tiers for mechanical fan-out.
Picking the tier is part of the architecture
◆ boundaries
agents work on development projects outside regulated environments.
Inside them, AI stays out of the critical path
email ▸ highmarck@member.fsf.org
github ▸ github.com/HighMarck17
fsf ▸ Free Software Foundation associate member