Backend & software engineer · Kragujevac, Serbia

From messy problems
to working systems.

I'm Janko Štrkalj. I build backend systems, AI pipelines and IoT software that turn ambiguous, high-stakes problems into something people can act on, owning features end to end, from database and API design to testing and deployment.

See the work ↓ Download CV
janko / current_focus.yaml
# where software meets the real world
role: backend & software engineer
stack: [Python, FastAPI, PHP, MySQL]
current_builds: [Riskform, GRC Assistant, Evidence Manager]
shipped: [Energy Orchestrator, GRCentric]
education: B.Eng. Software Engineering, 2026
method: frame → model → build → test
bias: explainable systems over black boxes
status: open to roles
Work

What I'm building, and what I've shipped.

Client systems I've delivered, plus private builds where AI helps people make better security and compliance decisions.

Shipped work

client projects · 2025–2026
Neighborhood Energy Orchestrator / IoT PoC · 03–06/2026

Coordinate the whole district, not just the house.

A proof of concept for a Dutch energy-tech client, part of an innovation project funded by the European Regional Development Fund (EFRO). It coordinates solar panels, home batteries, EV chargers and heat pumps across a neighborhood to reduce grid congestion. I built the edge software for two Raspberry Pi 4 devices, the FastAPI cloud orchestrator, and a simulation engine that models hourly per-household consumption.

≥15%lower transformer peak load
≥20%higher self-consumption
PythonFastAPIMQTTRaspberry PiInfluxDBSQLite
DELIVERED PoC / EFRO-FUNDED CLIENT PROJECT
EDGE2× Pi 4 · MQTT · device control · short-term predictionsense
↓
ORCHESTRATORFastAPI · simulation engine · weather & price feedsplan
↓
OUTCOMEKPI targets met · 8 & 30 houses · 100 / 700 kW limitsverify
GRCentric / Compliance platform · German-based startup · 2025

Compliance frameworks in, structured controls out.

A compliance-automation platform built end to end with a UI/UX designer: I covered the PHP/MySQL backend, database design and front end. A Python background worker picks up queued document jobs and feeds an AI pipeline that extracts security controls from framework documents and generates summaries. I moved it from a local Llama model to GPT-4.1 for better quality at acceptable cost.

PHPMySQLPythonGPT-4.1LLM pipeline
SHIPPED MVP / LATER PROMOTED BY THE CLIENT
INPUTframework documentsingest
↓
QUEUEPython worker · MySQL jobs · run lockprocess
↓
AI LAYERextract controls · summaries · Llama → GPT-4.1structure
↓
OUTPUTcompliance-automation platformact

Current builds

private · not public yet
Riskform / Risk Reduction Platform

Stop prioritizing vulnerabilities. Prioritize the work.

Riskform turns fragmented vulnerability and threat intelligence into explainable remediation campaigns, ordered by the risk they can remove and the effort required to remove it.

public data firstpolicy as codeexplainable decisionscampaigns, not CVEs
PRIVATE BUILD / NOT PUBLIC YET
INPUTNVD · CVSS · EPSS · KEV · threat intelsignals
↓
POLICYtaxonomy · scoring · grouping · planningexplain
↓
OUTPUTremediation campaignsnow / next / later
GRC Assistant / Private Build

Put AI inside the workflow, not on top of it.

GRC Assistant explores how cybersecurity maturity, assessments, controls, evidence, risk and audit readiness can operate as one connected system, with humans retaining decision authority.

human in the loopstructured evidencecompliance as codeAI-native workflows
PRIVATE BUILD / NOT PUBLIC YET
MODELMaturity → Assessment → Controlsstructure
↓
EVIDENCEdocuments · validation · linked riskscontext
↓
DECISIONrisk · mitigation · audit readinesshuman-led
Evidence Manager / Private Build

Turn scattered files into connected, searchable context.

A private, AI-assisted workspace that organizes files, notes, references and saved material into logical objects that can be searched, connected and revisited by meaning instead of folder location.

local-firstsemantic searchAI-assisted linkingpersonal knowledge
PRIVATE BUILD / NOT PUBLIC YET
INPUTfiles · notes · links · screenshots · referencescollect
↓
AI LAYERextract · relate · cluster · describeconnect
↓
OBJECTsearchable evidence with contextretrieve

Earlier projects

personal & university · open source
Graduation Project · 2026

Haiku Clock

An e-ink display that shows a haiku matched to the time of day, with day and night artwork. Syncs time over NTP, reads haiku off an SD card, and sleeps between 15-minute refreshes to save power.

MicroPythonRaspberry Pi PicoE-Ink
View on GitHub →
Personal Project · 2025

Music Recommendation

Suggests five similar tracks from audio features like energy, valence and danceability using KNN with cosine similarity, with a 2D PCA plot of where a song and its matches sit in feature space.

PythonStreamlitscikit-learn
View on GitHub →
Team Project · 2023

Travel Agency

A booking platform for summer destinations on every continent, with destination pages, reservations and automated email notifications. Built with a four-person university team.

PHPLaravelMySQL
View on GitHub →
How I work

I like the part before the obvious solution exists.

The useful work is rarely another dashboard. It's finding the real constraint under the request, making the logic explicit in code and data, and shipping a system that survives contact with reality.

01 / FRAME

Find the actual problem.

Separate symptoms from constraints. Map the data, the stakeholders, and where human judgment still has to stay in the loop.

Energy Orchestrator: the constraint was the transformer (100 kW / 700 kW limits), not any single device.

02 / MODEL

Make the logic inspectable.

Turn tacit reasoning into schemas, states, policies and pipelines that can be tested, challenged and improved.

GRCentric: document jobs go through a queued Python worker with a lock against concurrent runs, so the AI pipeline runs predictably.

03 / SHIP

Build the smallest credible system.

Get something real in front of users, then measure whether it actually moves the number that matters.

Energy Orchestrator: judged on KPI targets in 8- and 30-house simulations, not demos.

Toolbox

Skills

Languages & Frameworks
Python (FastAPI)PHP (Laravel)JavaScriptHTML/CSSStreamlit
Data
MySQLSQLiteInfluxDB
IoT & Embedded
MQTTRaspberry PiEmbedded Systems
AI & ML
LLM APIsscikit-learn
Backend & Tools
REST APIsGitLinuxWindows
Spoken Languages
SerbianEnglish
Career

Experience

Neighborhood Energy Orchestrator · Freelance, Dutch energy-tech client

03/2026 – 06/2026 · Remote
  • Built an IoT proof of concept for a neighborhood energy-orchestration platform, part of an innovation project funded by the European Regional Development Fund (EFRO).
  • Developed edge software for two Raspberry Pi 4 devices and a FastAPI cloud orchestrator that coordinates solar panels, batteries, EV chargers and heat pumps to reduce grid congestion.
  • Designed a simulation engine modelling hourly per-household consumption for 8- and 30-house districts under 100 kW and 700 kW transformer limits; with the orchestrator enabled, the system met the KPI targets of ≥15% lower transformer peak load and ≥20% higher self-consumption.

GRCentric · Freelance, German-based startup

07/2025 – 12/2025 · Remote
  • Built a full-stack compliance-automation platform end to end in collaboration with a UI/UX designer, covering the PHP/MySQL backend, database design and front-end implementation; shipped an MVP the client went on to promote to customers.
  • Developed a Python background worker that processes queued document jobs from MySQL, with a lock against concurrent runs, connecting the platform to an AI pipeline that extracts security controls from framework documents and generates summaries; moved from a local Llama model to GPT-4.1 for better quality at acceptable cost.

IT Support · Eurobank Direktna

12/2023 – 03/2025 · Kragujevac, Serbia
  • Day-to-day technical support for internal bank teams: software installation and troubleshooting of workstations, printers and banking systems.
  • Handled support requests through Jira, diagnosing and resolving hardware, software and application issues for bank staff.
  • Wrote SQL queries to pull data from internal databases for requesting teams.

Internship · Comtrade – EDIT

07/2023 · Kragujevac, Serbia

Two-week team internship: designed and deployed a cloud-native application.

Education
Bachelor of Engineering, Software Engineering · Faculty of Engineering, Kragujevac · 10/2019 – 09/2026
Specialized Program in Mathematics · First Kragujevac Gymnasium · 09/2013 – 06/2019
Organizations
Youth Club SKC · member, student cultural activities and events · 02/2025 – present
Neutrinos IT · member, app development and local community projects · 01/2025 – present
Get in touch

Working on a backend, AI or IoT system?

I'm open to backend, full-stack and IoT-leaning roles, and happy to talk about anything messy, cross-functional and hard to make operational, especially if it involves gluing hardware to software. Email is the fastest way to reach me.

Connect on LinkedIn ↗ GitHub ↗ Download CV