Engineering · AI Security · Management

Karim Karimov

I build and secure the systems machine learning runs on, and manage the processes that move them forward.
Three years of production engineering at Sberbank-Service, the technology arm of Russia's largest bank: load-testing AI inference platforms and automating quality for five products. Now completing a dual master’s in innovation management and business administration at HSE and UNICAMP.
Resume
Industry
↓
CV
Academic
↓
Open to remote work and relocation
Campinas, Brazil
3 yrs
production engineering at Sberbank-Service
5
production products covered by my test automation
2
master’s programmes in management, in two countries
4
languages: Russian, English, Portuguese, German

Approach. Progress, straight ahead.

I’m an engineer by training and a manager by choice.
Three years of building production systems taught me that the hardest problems sit above the code — in how processes and ownership are designed. AI security is exactly that kind of problem, and the structure to handle it doesn’t exist yet. I intend to build it.
UNICAMP · Limeira campus, at sunset
01
Build.
Three years shipping and testing production systems: test automation across five products, load testing for AI inference, and CI that catches regressions on every merge.
TypeScript
Python
Locust
GitLab CI
02
Secure.
A degree in information security and research on AI in cyber defence, from on-device vulnerability triage to how AI adoption concentrates risk across whole markets.
OWASP
MITRE ATT&CK
CVSS
LLMs
03
Lead.
Two master’s programmes in innovation management and business administration: R&D and innovation strategy, IP management, and structured, evidence-based decisions.
Innovation management
IP strategy
Decision analysis
The path so far
▼
2021
Information Security
HSE · BSc
2023
Software Engineer
Sberbank-Service
2025
Innovation Management
HSE · MSc
2026
Business Administration
UNICAMP · MSc
Next
Engineering × management
in AI security

Work & Research. Where reliability, security and machine learning meet.

Current project
Apr 2026 – present
AI Pentest Tool
LLM-assisted vulnerability triage that runs entirely on-device. It orchestrates seven open-source scanners and uses a locally hosted model to correlate their output into one deduplicated report, with CVSS 3.1 vectors, OWASP Top 10 and MITRE ATT&CK mapping, and concrete remediation steps.
Scan data never leaves the operator’s machine. For security-sensitive assessments, that rules out hosted models.
Python
Ollama
pytest
View on GitHub ↗
1
Scan
Nmap
Nuclei
Nikto
Gobuster
testssl.sh
subfinder
HTTP headers
raw findings
2
Correlate
Local LLM · Ollama
dedupe · score · map · advise
one report
3
Report
CVSS 3.1
OWASP Top 10
MITRE ATT&CK
Remediation
Runs on-device. No data leaves the machine.
Master’s thesis · UNICAMP
From Reactive to Proactive Cybersecurity: The Role of Generative AI and Large Language Models in Predictive Threat Management
Advisor: Izabela S. Rampasso · expected 2027
How generative AI can move security teams from responding to incidents to anticipating them.
Term paper · HSE · 2026
Global Implications of AI-Powered Solutions for Cybersecurity in Companies
Supervisor: Prof. Konstantin Vishnevskiy
A comparative study of the EU, US and Russia built on nine expert interviews. Rational, firm-level AI adoption adds up to layered market concentration, regulatory fragmentation and a governance gap no single actor owns.
Publication · Aug 2026
Public Perception of Innovation: Good for Society or Made by the Devil?
The Social Life of Innovation, HSE Basic Research Program Working Papers, WP BRP 130/STI/2026, pp. 70–75
Chapter in an edited collection with contributors from HSE, Tsinghua, BUPT and UNICAMP. It argues for literacy and oversight, not resistance, as the response to technological harm.
Read on SSRN ↗
Selected projects. Research, tooling and teaching.
HSE MIEM · 2025
↗
MLOps Execution and Monitoring System
Automated API and web-interface testing for a platform that deploys and monitors production ML models, including coverage for model-serving endpoints.
Research practicum · 2023
↗
Web Attack Detection with CNN-BiLSTM
Reproduced a CNN-BiLSTM intrusion detector on CIC-IDS2017. A Random Forest baseline beat it on recall and F1, and the deep model nearly failed on the rare XSS and SQL-injection classes.
HSE MIEM · 2024–25
↗
Systems & Networks E-Learning Platform
Authored 20+ lab assignments for an undergraduate course and built the web interface of its delivery platform.
Study project · 2022
↗
Cryptosystems in Python
Classical and modern cryptographic primitives, implemented to study how they fail.
Experience
Software Engineer
Sberbank-Service · Moscow
Jun 2023 – Jul 2026
Built automated test suites in TypeScript, Jest and Appium across web, mobile and API layers for five production products.
Designed load-testing scenarios for backend and AI inference platforms with Locust and Python, surfacing bottlenecks before release.
Wired the suite into GitLab CI so regressions were caught on every merge.
Traced failures across distributed services with Grafana and the ELK stack; validated data integrity in production PostgreSQL for core banking workflows.
Education
UNICAMP
2026 – 2027
Master, Business Administration (dual degree)
Campinas, Brazil
HSE University
2025 – 2027
Master, Science, Technology & Innovation Management and Policy
Moscow, Russia
HSE University
2021 – 2025
Bachelor, Information Security
Moscow, Russia

Now. What I’m focused on this season.

Updated October 2026
Researching
How LLMs can move security teams from reactive response to predictive threat management. That is my thesis at UNICAMP.
Studying
Innovation and R&D management, IP strategy and decision analysis: how technology gets steered, not only built.
Building
The AI Pentest Tool: better correlation, fewer false positives, fully on-device.
Living
In Campinas, Brazil, on a dual-degree exchange, and getting my Portuguese past A2.
Preparing
TOEFL in October 2026.
Looking for
Roles where engineering meets management in AI security: research and security engineering, technical programme management. Remote or with relocation.

Personal. The person behind the résumé.

Shot on film
Moscow → Campinas
Two countries, one degree.
I moved from Moscow to Brazil for a dual master's between HSE and UNICAMP: two universities, two academic cultures, one thesis.
≈ 11,800 km
Moscow
55.76° N · HSE University
Campinas
22.91° S · UNICAMP
Teaching
20+ labs, written from scratch.
While still a student, I designed the practical assignments for an undergraduate systems and networks course and built the platform it runs on.
Research habits
I report what didn’t work.
When a simple Random Forest beat the deep model I was reproducing, that became the result. Honest baselines matter more than impressive numbers.
Technology & society
Beyond the code.
My published chapter is about how people perceive innovation, not how to build it. Securing AI is as much a governance problem as an engineering one.
Languages
Four, and counting.
Russian
Native
English
C1
Portuguese
A2
German
A2
Contact

Let’s talk.

Research collaborations, roles where engineering meets management, or a question about my work. Email is the best way to reach me.
Email
mirakky99@gmail.com
↗
Telegram
@mirakky
↗
GitHub
github.com/mirakky
↗
Campinas, Brazil · Open to remote work and relocation
© 2026 Karim Karimov
Moscow → Campinas → Straight Ahead