About
I've been building software for about twelve years, remotely, for teams across the US and Europe. These days I take products from idea to production and keep them running: the architecture, the code, the infrastructure, and the incidents. Mostly TypeScript, Go and Python, on AWS and Kubernetes.
Recent work covers IoT time-series pipelines on ClickHouse, AI features that real customers use (RAG, tool-using agents, MCP), cloud security tooling, and EU Cyber Resilience Act compliance. I like owning a system from the first sketch to the 3 a.m. incident report, and working directly with founders and non-technical people.
I'm open to senior full-time remote roles or B2B contracts. Say hi at helloismajl@gmail.com.
Work
-
Lead software engineer for a Danish maker of industrial heat pumps (110 kW to 2.5 MW). I built and run essentially their whole software platform: the AWS IoT and serverless backend, a ClickHouse time-series pipeline, the portal where staff, partners and end users monitor and manage the heat pumps, a webshop, an ops console, an AI knowledge base, and their EU CRA compliance tooling with signed SBOMs.
-
Cybee.ai2024 - 2026
Senior backend engineer on a cloud-security / EDR SaaS. Node.js and Fastify services with MQTT workers feeding MongoDB and Elasticsearch; Cloud Guard, a customer-deployed Go agent collecting audit logs from Google Workspace, Microsoft 365, GitHub and Slack; and "Cybee CISO", an in-product AI assistant with 14+ tools over each customer's security data.
-
Before that2014 - 2023
Remote contracts and roles with startups in the US and the Nordics:
- Butler Hospitality: lead engineer and chapter lead. Mentored squad leads, built the shared React component library and the ordering services on the main revenue path.
- Polymaths: HubSync, a tax platform for CPAs, and React Native apps for the U.S. Anti-Doping Agency.
- Ornio: dashboards and Node.js services for Tipio.no, a group-buying marketplace, plus a localisation system used across all apps.
- Vinden.se, OpenCybernetics, Appins.io: AWS backends, a Web3 trading dapp, and a no-code builder for internal tools.
- Freelance: a website and online-store builder behind 2,000+ sites, a vertical search engine, and coaching six summer interns.
- iziSurvey: my first job, on a survey tool for researchers.
Projects
-
A business platform that SMBs in Kosovo use to run their day-to-day operations. I built the web app, the mobile apps, the point of sale, the tax-authority integration, and an AI assistant that turns receipts and invoices into draft entries.
-
An AWS cloud security posture scanner in Go. 24 scanners and 83 Starlark rules across 19 AWS services, real-time CloudTrail ingestion over SQS, finding lifecycle and SLAs, Slack and PagerDuty alerts.
-
Thermonova knowledge base
An internal AI assistant over refrigeration standards: graph and vector RAG (Neo4j, Qdrant), a FastAPI service streaming Claude on AWS Bedrock, and an MCP server. Every answer cites the exact clause.
-
CRA compliance toolset
EU Cyber Resilience Act tooling built in-house: product registry, signed SBOMs, Dependency-Track triage, Article 14 reporting against CISA KEV and ENISA feeds, a PSIRT inbox and a locked evidence store.
Pet projects
-
My own programming language in Go: an interpreter with async/await, a type checker and a standard library written in the language itself, then a bytecode VM, then native code through Cranelift and LLVM.
-
A coding agent in Python built from scratch on Gemini: tool calling, streamed reasoning, sessions, plan/execute mode and a Textual TUI.
-
An eBPF process monitor in Go, written to learn eBPF: hooks kernel tracepoints for exec, exit, network connections and sensitive file access, with a TUI and a web view.
-
Dart Engine
An open API protocol with a Go implementation for building APIs quickly.
-
Img Server
A Go image server: upload through pre-signed URLs, then resize and transform by changing URL parameters.
-
Rubiv.io
A canvas editor that generates banners by mapping content into design templates.
Misc
- I learn a domain before I build for it. For Thermonova I wrote myself a heat-pump physics study path tied to the platform's own variables.
- I write the plan before the code: what I found, the options and trade-offs, how to run and roll back, and why we chose what we chose.
- AI made writing code cheap. Knowing what to build, and understanding every line it writes for you, got more valuable.
- I tinker with hardware when I get the chance: microcontrollers, Raspberry Pi, IoT gateways and OpenPLC. Still a beginner, but it helps to have held the other end of the MQTT connection.
- Studied Computer Science at Universum College, Pristina.
- Albanian (native), English (professional).