D·B
Dragan Basta
Engineering Leader · Product, Growth & AI · Belgrade · CET

Dragan Basta

I grow engineers and ship products people enjoy using.

Fourteen years in software, eight of them building the teams that write it. I've led at every level: coaching engineers one-to-one, managing managers in a 30-person org, running technology as CTO, and leading a company I co-founded through an acquisition. Most of that work sits where engineering meets growth: getting people to their first win, to paying, and to staying. I still write code every week.

3 → 25Remote team built across three continents
0 → 30Engineering org co-founded, with managers reporting to me
10,000+Consumer devices shipped
OneAcquisition led, from due diligence to integration
A note on leadership

What I've learned building teams

I build teams that ship. At PeakNick I grew engineering from zero to thirty people, with managers reporting to me. At Gripbeats I grew a team of hardware and software engineers from three to twenty-five, across three continents.

Good teams aren't built in the annual review. They're built in the weekly one-to-one, in feedback given the same week, and in a decision written down so no one has to wait for me.

I still write code. It keeps my estimates honest and keeps me close to the team. Today that includes building with AI: clear rules, small batches, tests, and a person who signs off.

I'm looking for my next team, leading engineers directly or leading the leaders, on a product that cares about craft and writes things down.

Dragan BastaBelgrade, November 2026

Six principles I lead by

Coaching is a schedule

A weekly one-to-one that belongs to the engineer, and a written growth plan we both keep current.

Feedback early, in writing

Feedback given the same week is cheap. Saved for the review, it's expensive.

Writing makes remote work

Decisions live in short docs with an owner, so nobody eight time zones away waits for me.

Stay close to the code

I review pull requests and prototype. It keeps my estimates honest.

Ship, measure, decide

Polish or ship now is an open call with Product and Design, written down.

Raise the bar for everyone

Light standards for review, delivery and hiring, light enough that people actually use them.

In practice
SwishGuru, at PeakNick

Math vs algorithm first

A mathematician on our SwishGuru team, working on shot analysis, thought math-first, not algorithm-first. We lost days debating the theory behind each idea, which a startup can't afford.

So before any development I started async one-to-ones with him: first the system architecture and the expected inputs and outputs, then the maths. After a month we no longer needed them. We delivered the architecture, he polished it, then did the maths. By month three every algorithm change came through without long meetings, because he documented the path from architecture to final proposal himself. The whole team felt it: we ran experiments faster, deployed to testing faster, and sat in fewer meetings.

Mentoring

From quiet to Tech Lead

One of our engineers was shy. He avoided meetings and rarely spoke in them, but he was passionate about frontend development and network performance.

So I started Tech Talks, where anyone on the team could present what they cared about. He took his turn and taught the rest of us about performance and network issues. Then he went further: he fixed them across our projects and took responsibility for that work. He eventually became our Frontend Tech Lead.

Performance · PeakNick

Beyond the happy path

A backend engineer was shipping code that only covered the happy path, and he kept ending up in conflicts with teammates. I saw the first in his pull requests and heard the second in their complaints.

I talked to him directly. For the code, we broke every task into tiny pieces together, so each one was clearly finished; after a while he did the breakdown himself and I reviewed it. For the conflicts, I talked to everyone involved and gave each person exact responsibilities and my expectations. At first his work deliberately didn't overlap with anyone else's, then we slowly added the overlap back. I passed every complaint on to him plainly, because I couldn't put him in front of a client until it was solved. It took a lot of time, but he improved.

What I learned: a good storyteller with real knowledge can still miss the details. And someone polite with their manager can still be close to a conflict, so I read the written conversation in Slack and Jira too. When someone gets stuck, avoids others or goes silent, that's a signal, and usually not a good one.

Staze Srbije

Agents refactor, people decide

Staze Srbije is Serbia's national hiking-trail registry: legacy code, integration with ministry servers, and GIS data from GDi. It needed fixing, so I put a team together.

I led that team through a switch to working with AI agents: the agents do the hands-on refactoring in small pull requests, and the team reviews and merges each one, so a person signs off on every change. I kept the client in the loop the whole way. Together we fixed the platform and removed more than 500 duplicated language blocks.

Selected work

Teams I've led and what I'm building now.

Leading teams
Shipped · 10,000+ units

Gripbeats

An AI bracelet that turns movement and touch into music. I built the remote team of hardware and software engineers from 3 to 25 people across three continents and led the technology from firmware to cloud, through 10,000+ units in customers' hands.

Chief Technology Officer · 2022 – 2025

Co-founded · AI shooting coach

SwishGuru / Ball Skiller

A personalized basketball shooting coach in your phone. Film your shot, and the app tracks your form in 73 points, detects mistakes and builds a training program around them. I co-founded it as CTO and built it with PeakNick's engineering team.

Co-founder & CTO (Ball Skiller), Head of Engineering at PeakNick · 2020 – 2025

Advising now
Part-time · early demo

NaraAI

I'm the technical advisor to NaraAI's two founders, and the only technical person on their team. It writes regulatory-grade patient safety narratives from clinical-trial data. The model writes the prose and deterministic code owns every fact. A 14-check verifier and evals gate every release. On the public pilot study it already flagged all 3 fatal events miscoded as “not serious”.

Technical advisor, part-time · 2026 –

Leading at scale

Beyond one team: how I've shaped orgs, hired across time zones, reported up and led a company through an acquisition.

PeakNick

Org design

I split engineering by platform, so each of my 2–3 engineering managers led the stack they knew best. Before that, a manager strong in DevOps spent a lot of time in meetings with one strong in mobile just to settle everyday questions. The split took most of those meetings away. From 2022 the managers ran engineering day to day without me, and I moved into a director role: high-level decisions, budget, and advice on hard technical calls.

Gripbeats · 3 → 25, three continents

Hiring at scale

We grew with the same steps for everyone: a screening call, then a take-home task or a one-to-one problem-solving interview, with no live coding. I looked for good people who communicate well, and I made sure every candidate reached a moment where the honest answer was “I don't know”, to see how they handled the pressure. I hired for people who stay, and they did.

Getting people to the good part

My growth work has been as a founder and builder: the first result, the first minute, and coming back the next day.

Activation

SwishGuru

5 → <2 minwait for your first shot analysis

We made one moment the activation goal: seeing your own shot drawn in 73 body keypoints, with your mistakes named. Onboarding was cut to three steps to get there: download the app for free, film your shot, upload it.

The first version took about five minutes to analyse a shot. We moved the pose model from CPU to GPU instances on AWS, made the algorithm close to 60% faster, and got the wait under two minutes. While players waited, the app showed basketball tips and fun facts, so the wait felt like part of the app, not a loading screen.

We were strict about what to leave out. No sign-up, profile or tutorial before your first analysis: the account came after the “aha”, not before it. We also chose not to run the analysis on the phone. On-device would have felt instant, but on most players' phones it was too slow and not reliable enough, so we kept it in the cloud and made the wait worth watching.

Scale

Gripbeats

10,000+units in customers' hands

For a consumer device the first minute decides everything: put the bracelet on, pair it, move your hand, hear music. Engineering's job, across firmware, sensor fusion, BLE and the apps, was to make that minute work every time.

Users who stay

Basta Crystal

15 users5 companies run their invoicing on it

It started as an invoice and offer tool for a few friends' companies. They kept using it and asked for more, so my team grew it into an all-in-one workspace for invoices, tasks and saved pages. Small, but real: our users open it every day, and could switch tools any day, but haven't.

Building with AI, carefully

AI writes a lot of my code, and every repo I own follows the same four steps. It's the practice I'd bring to a team, so engineers can move faster without lowering the bar. Since 2025 I've also set it up for about a dozen small and medium-sized companies, mostly in health tech.

Write the rules down

Every repo has a CLAUDE.md with the rules that never bend and what to read first.

Work in small batches

An agent does the hands-on work in small pull requests that are easy to review.

Make mistakes testable

Every bug starts as a failing test, and evals in CI can block a release.

Keep a person in charge

Schemas, data, money and patients need a human sign-off. The agent asks, it doesn't guess.

Dragan Basta, by the sea

About me

I'm an engineer from Belgrade who ended up building teams, and never stopped building software.

I studied Software Engineering at the University of Belgrade's School of Electrical Engineering, earned a BSc and an MSc there, and taught there as a teaching assistant. Since then I've worked across games at Nordeus, AR and computer vision at Wolf3D, consumer hardware at Gripbeats, and health tech. I co-founded PeakNick and led it through its acquisition.

Today I help companies build with AI, advise NaraAI, and run a few small products of my own. Along the way I write down what I learn about leading teams that build with AI.

Based in
Belgrade, Serbia · CET
Works
Remote-first since 2018
Languages
English, Serbian
Education
BSc & MSc, Software Engineering
Off the clock
I sail, ski and play Lego!

Experience

  1. 2025 – now

    Independent

    AI engineering advisor & fractional CTO

    I help small and medium-sized companies, about a dozen so far and mostly in health tech, build with AI: agent workflows, integrations between their tools with checks that catch human errors, and evals. The result is fewer bugs and fewer meetings, because the work is written down. For a winemaking company I set up daily business and financial reports, so everyone knows what is where in the pipeline. Alongside that I advise NaraAI and maintain TValuer and Staze Srbije. Side projects: Basta Crystal, PixelTest QA and Basta Jumper.

  2. 2022 – 2025

    Gripbeats

    Chief Technology Officer · full-time from 2022

    Built a fully remote team of hardware and software engineers from 3 to 25 across three continents. Led BLE, embedded, AWS, mobile and AI/ML, shipped 10,000+ units, and reported to the CEO and investors.

  3. 2018 – 2024

    PeakNick

    Co-founder & Head of Engineering

    Built a 30-person engineering org with 2–3 engineering managers reporting to me. Ran a portfolio of products including Ball Skiller / SwishGuru, TValuer and Staze Srbije. From 2022 my engineering managers ran day-to-day work while I was full-time at Gripbeats. I stayed on as co-founder and led the company through its acquisition in August 2024, from due diligence to integration.

  4. 2019 – 2020

    PatchAI

    Fractional CTO · part-time, alongside PeakNick

    Led two distributed vendor teams building a clinical-trial platform, migrated it from IBM Cloud to AWS, and delivered the MVP on time ahead of Series A validation.

  5. 2012 – 2019

    Earlier

    The Local Europe · Wolf3D · Nordeus

    iOS Tech Lead, AR / computer-vision engineer, and software engineer. Alongside that, a teaching assistant at the University of Belgrade School of Electrical Engineering, where I earned a BSc and an MSc in Software Engineering.

Side projects

Small products I build and run, to stay close to product and craft.

Side project · 15 users, 5 companies invoicing · 370+ commits since April 2026

Basta Crystal

It started as an invoice and offer tool for a few friends' companies. My team grew it into an all-in-one workspace for business, tasks, saved pages and a daily practice, on web, iPhone, iPad and Mac. Our users open it every day.

Read the case study
Side project · built for one product designer

PixelTest QA

Design review that compares a Figma design with the built frontend and turns “this feels off” into a list of specific differences that a designer and an engineer can agree on.

Read the case study