MKDownload CV

MATVEI
KANDALINTSEV

Break it. Understand it. Build it better.

#1in the world8,073requests. Zero errors.CTOat nineteen<500msto stop a live attack+61%over baseline8.42/10honours GPA3native languages17age at first paying client0animation libraries on this site

[ 01 ] The arsenal — selected systems

Pick a system. Watch it run.

Every card below runs its own simulation — a hand-coded canvas program, written for this page, that shows what the real thing does. No video, no animation library. Hover to bring one online; open it for the full case file.

01Break

I'm a certified penetration tester — trained to hack systems legally — and I compete on a hacking team. I find out how things fail before attackers do.

02Understand

I research how AI reads the architecture of software, and I've built machine learning from raw mathematics — no libraries — to know exactly what's inside the box.

03Build & prove

I'm the CTO and lead developer of real platforms, and every claim I make ships with evidence: stress tests, security audits, isolation tests, world rankings.

Open file ⏎

An AI platform that gives CISOs evidence-backed due diligence before they trust a security vendor with millions — architected so unverified evidence can never change a verdict.

8,073 requests, zero errorsAI engineering · Security
Open file ⏎

CTO at 19 of an AI admissions platform that answers students from source-verified official data — with data isolation and auth hardening enforced in the architecture, not bolted on.

60+ isolation tests, live DBAI engineering · Security
Open file ⏎

A defence against USB sticks that pretend to be keyboards — it recognises non-human typing and locks the machine in under half a second.

<500ms detection latencySecurity · AI engineering
Open file ⏎

A global trading competition against 19,000+ teams. Finished in the top 1.3% — and ranked #1 in the world in the manual round.

#1 worldwide, manual roundData analysis · AI engineering
Open file ⏎

A self-serve SaaS that turns any URL into an AI audit in minutes — and since it fetches arbitrary user-supplied URLs, it was built security-first from the first line.

SSRF defended by designAI engineering · Security
Open file ⏎

Machine learning built from raw mathematics in Java — no AI libraries allowed — predicting student grades from 16,994 real records.

16,994 real recordsAI engineering · Data analysis
Open file ⏎

A red-team harness that attacks an LLM with known manipulation patterns and has a second model judge every response SAFE, BORDERLINE or UNSAFE — repeatable risk you can track release to release.

SAFE/UNSAFE judged per responseSecurity · AI engineering
Open file ⏎

A consumer mobile app for mass-deleting your old Instagram messages — private by design: message text never leaves your phone.

On-device zero data leavesFull-stack
Open file ⏎

I build the software that runs a business: custom CRM systems that capture and follow up every lead, and websites engineered to convert — delivered directly, no agency in between.

17 age at first paying clientFull-stack

[ 02 ] Dossier — subject file

I attack systems first, then build them to hold.

I'm 19, in my second year of AI & Data Science at Maastricht University. I hold the eAIS and eJPT security certifications, research LLM-driven software-architecture extraction, and lead the technical side of real AI platforms. I approach every system from the attacker's side first, then build it so that security, reliability and evidence integrity are properties of the architecture — not checklists bolted on at the end.

Attack from one side. Defend from the other. Know both.

Currently

Lead AI Developer

VendorVerdict

AI cybersecurity vendor-intelligence platform

2026 — present

Chief Technology Officer

AdmissionFit

AI-driven university admissions platform

2026 — present

Undergraduate Researcher

Maastricht University

MaRBLe Honours · LLM architecture extraction

2026 — present

CTF Team

MaaSec

web exploitation · reversing · crypto · forensics · OSINT

2026 — present

Four domains, one method

AI engineering

I make AI systems that work in production, not just in demos

VendorVerdict · SiteScore · LLM safety harness · University research

Security

I'm certified to break into systems — and to defend them

Penetration tester (eJPT) · AI security specialist (eAIS) · DuckyFence · Competitive hacking team

Full-stack engineering

I own entire products end to end: database, backend, interface, deploy

CTO at AdmissionFit · VendorVerdict stack · SiteScore · Client CRM systems

Data analysis

I apply machine learning and statistics under real pressure

IMC top 1.3% worldwide · ML from scratch in Java · 30% trading-model uplift

[ 03 ] University research

Teaching AI to read the architecture nobody wrote down.

A system's architecture usually lives only in its code — documentation rots, intent gets lost. I research whether LLMs can recover that hidden structure directly from source, and whether architecture-aware models write safer, more correct code when translating between languages. It sits exactly where my two interests meet: AI, and the security of what AI builds.

TRACK 01

Architecture recovery

Can a model reconstruct a system's structure, dependencies and design intent from raw source alone?

TRACK 02

Cross-language translation

Does architecture-awareness make AI better at rewriting software from one language to another?

TRACK 03

Security of AI-written code

Where does AI-generated software fail — and does recovered architecture predict it?

Maastricht University — Dept. of Advanced Computing Sciences (DACS)

MaRBLe Honours Research Programme (selective, research-track)

Supervised by Dr. Ashish Sai · May 2026 — present

Source code in→architecture out

live extraction · illustrative

[ 04 ] Contact

Selectively open.

Capacity is limited on purpose. The interesting work gets everything I have — the rest gets a polite no. If it's ambitious, technically hard, and someone has to own it end to end, it belongs in my inbox.

01

Work with me

High-ownership roles and hard technical problems. I take very few at a time, and only the ones worth doing properly.

  • AI systems and security engineering, owned end to end
  • Research collaborations
  • Technical co-founding / CTO-level ownership
  • A focused product build for your business
Make the case →

02

Own the products

Finished builds you can buy the rights to, license, or grow with me as an ambassador.

  • Retractmobile · bulk Instagram unsending
  • DuckyFenceUSB attack detector
  • LLM safety harnessAI manipulation testing
  • Grade-prediction MLfrom-scratch machine learning
Buy rights / become an ambassador →

Direct line

contact@kandalintsev.com
Download CV (.docx)LinkedIn ↗

Serious inquiries read same-day · Maastricht, Netherlands