EN · RU
AI Engineer · Available for projects

Alexander
Kirilov

Telegram LinkedIn GitHub
Alexander Kirilov
scroll
Background

From high-load systems
to AI engineering

2006
Foundation in engineering
Started as a Java developer, built products across fintech, e-commerce, and logistics. Learned early that software only matters when it runs reliably at scale.
2012
Odnoklassniki – real scale
Joined one of the world's largest social networks: 140M users, 350K req/sec. Designed targeting and notification systems. Learned what production-grade really means.
2016
AI/ML at Wargaming's InData Labs
Led backend teams on ML-powered products. Integrated ML modules into high-load production services. First deep contact with AI systems in production – not theory, but real revenue impact.
2024
AI Engineer – full focus
LLMs, RAG pipelines, agentic systems. Building PsychoAI – an AI platform for the mental health space. 20 years of production experience now applied to what's next.
What I do

I build AI that ships

🧠
AI Integration
I connect LLMs, RAG pipelines, and agentic systems to your existing business processes – so the result is working software, not a proof of concept.
🏗️
Architecture to Production
From system design to Kubernetes and cloud deployment – I own the full delivery. Monitoring, caching, error handling, CI/CD. Ready to hand over, not to babysit.
💬
Business Communication
I speak both languages – technical and business. I clarify what's unstated, surface what's missing, and build what actually solves the problem – not just what was asked.
👥
Team & Process
20 years of leading engineering teams. I set up Scrum, own estimations and risk, and can step in as a tech lead or hands-on developer – whatever the project needs.
What I'm building

Products in development

In development · 2026
PsychoAI

A neural network of psychology – one unified platform, two connected products. Between sessions, the system tracks the client's emotional dynamics and psychological wellbeing indicators. Before each session, the therapist sees how the client's emotional state evolved – mood shifts, key signals, overall tone. After sessions: topic analysis, long-term patterns – and intervention suggestions: the system highlights what's happening with the client and recommends what to try next. Accumulated cases and feedback on what worked feed collective intelligence – so the system can suggest the best approaches for each new client.

PsychoAI for Clients
  • Reflective diary – voice or text, daily
  • AI neuro-therapist – from quick check-ins to full sessions
  • Exercises, metaphorical cards, emotional tracking
  • Weekly reports with patterns and dynamics
  • Share diary and reports with your real therapist
data shared
PsychoAI for Therapists
  • Upload any session audio – transcribed automatically, gestalt patterns and key markers highlighted
  • Client card with long-term trend analysis
  • Pre-session briefing – what happened since last time
  • Pattern alerts – surfaces what's hard to notice manually
  • Diary shapes the emotional profile before a session – themes and tone
▶ Read more about the system
Client app – full feature set
  • AI sessions – choose a therapist persona: Perls, Rogers, Yalom or real practitioners
  • Metaphorical card work – AI-guided reflection around a generated image
  • Monthly reports with dynamics, topics, emotional timeline
  • Therapist marketplace – AI matches you to a practitioner by style and fit
Therapist app – full feature set
  • Intervention suggestions – what worked in similar cases across thousands of sessions
  • Client dynamics over time – topics shifting, emotional patterns, progress markers
  • Therapist notes by voice or text – auto-structured into a client epicrisis
  • Collective intelligence – a junior therapist gets the accumulated experience of the field
Privacy
  • Client controls what gets shared and with whom
  • Anonymisation before sending to cloud LLMs
  • Roadmap to fully local models – zero data leakage
Selected Work

AI systems built & shipped

psychoai-rag-assistant
Production RAG assistant for gestalt therapy. Faithfulness=1.0 evaluated with RAGAS. Python, Qdrant, Redis, Gemini Flash.
generic-rag-agent
Universal RAG platform – upload any documents, get a ready AI assistant. Python, FastAPI, Qdrant, Docker.
multimodal-telegram-bot
Telegram bot with session lifecycle – text, voice (Whisper STT) and photo (Gemini Vision) in one conversation.
ai-personal-assistant
Universal AI assistant – text, voice (Whisper + TTS), vision, image generation (DALL-E 3), RAG over your own documents.
Contact

Let's build something real

If you have an AI project in mind – or aren't sure yet what's possible – reach out. I'll come back with a clear picture of what can be built and how.

Write on Telegram Connect on LinkedIn