// olanokhin.com
Alex
Anokhin
Founding AI Engineer · Forward-Deployed AI
Production RAG · Agentic Systems · AI Security
Turning ambiguous business problems into adopted AI systems.
Currently @ Exyte · Heilbronn, Germany
93%
EA assistant
answer quality
answer quality
8×
LLM input reduction
1M → 115k rows
1M → 115k rows
4×
Meeting review
speedup @ Wire
speedup @ Wire
// about
I'm Alex — a founding-style AI engineer based in Heilbronn, Germany. I work best where the problem is underspecified: the business owner has pain, the data is messy, system boundaries are unclear, and someone needs to turn it into a production workflow.
My operating range is full lifecycle: discovery, solution design, evaluation, implementation, deployment, adoption, and stakeholder ownership. At Exyte I independently scoped and delivered AI initiatives across procurement, enterprise architecture, knowledge management, and governance. At Wire I was the sole AI engineer on a cross-functional note-taking initiative spanning product, mobile, backend, and security.
Before AI, I spent 10+ years in live technical production: concerts, broadcasts, corporate events, and protocol-controlled rooms where failures were public and recovery windows were measured in seconds. That is where my production mindset comes from.
Outside employment, I build research and open-source systems around vector search, agent security, trust infrastructure, and multi-agent review. I'm a 3x hackathon winner, accelerator winner, AI-startup feasibility judge, mentor, husband to Angelika, and father to Anna.
// operating mode
Embed First
Work directly with business owners and engineering teams until the real constraint is visible.
Define Evals Early
Build test sets, scoring methods, and traceability before scaling prompts or adding orchestration.
Ship Narrow
Scope the first useful workflow tightly, get adoption, then generalize from real usage.
Control Cost & Latency
Prefer deterministic filtering, caching, and HITL controls before spending tokens blindly.
// work
2026 — present
AI Engineer
Exyte Management GmbH · Stuttgart, Germany
- Embedded directly with business owners and enterprise architects as de-facto owner of AI initiatives across enterprise architecture, procurement, knowledge management, and governance.
- Enterprise Architecture AI Agent: designed, built, and evaluated in 4 weeks; achieved 93% answer quality and 87% context-match in single-shot mode; deployed via Microsoft Copilot Studio.
- Procurement Data Enrichment: reduced LLM input volume ~8x from 1M rows to ~115k via deterministic non-Latin filtering, cross-session caching, glossary controls, and HITL refinement.
- RFI Knowledge Management: replaced manual SharePoint/Excel search with hybrid RAG — BM25 + semantic retrieval + LLM reranker — with eval datasets, editing, versioning, and authorship traceability.
- AI Governance Intake: replaced a ~1-week manual scoring cycle with automated prioritization, instant notifications, and auto-created tickets via Microsoft Forms, SharePoint, and Power Automate.
CausLock
Governance Protocol · Human+AI Collaboration · In Progress
- Human+AI collaboration governance protocol with ticket-as-spine architecture linking Linear, GitHub, Slack, and Notion into an engineering knowledge graph.
- Introduces Bus Factor Score, AI Slop Score, and Legacy Score metrics to track team resilience, AI-generated noise, and long-term maintainability.
- Motivated by a concrete review asymmetry: Git was designed for human collaboration at roughly 100 words/minute, while AI agents operate at 200+ tokens/second, making traditional code review physically impossible at agent speed.
2025
AI Integration Engineer
Wire Germany GmbH · Berlin (Remote) · Apr – Sep 2025
- Built a working AI note-taking PoC in a single day and pitched it at company-wide sprint planning to ~100 people including the CTO; greenlit for production.
- Sole AI engineer in a 3-month cross-functional implementation with Product, Design, Android, iOS, Backend, and Security; reached internal staging.
- Cut meeting review time ~4x via transcription + summarization service under strict security constraints, including custom-compiled FFmpeg and stripped I/O.
- Built a Reminder Bot with custom NLP-to-cron parser and interactive button UI, replacing rigid slash commands for non-technical users.
KIWI Knowhow
Co-founder & Tech Lead · Virtual Senior Engineer RAG Platform
- Multimodal RAG for engineering knowledge retrieval across text and diagrams from technical PDFs.
- Won the CyberValley AI Founders pitch competition in the Stuttgart/Tuebingen accelerator ecosystem, earning 1st place and EUR 10,000 in AWS credits.
2024
ClyrAI
Freelance · AI Content Strategy Agent
- LinkedIn brand-health agent: custom parser → AI agent → ready PDF report.
- Cut reporting time ~3x for SMM workflows.
Abbey Tholey — Virtual Monk
Freelance · Context-Aware RAG Agent
- Digitized museum archives dating back to 634 AD — nearly 1,400 years of history — into an interactive RAG assistant with persistent session memory.
GreeterAI
Co-founder & Solo Developer · Generative Video Greetings
- End-to-end generative video pipeline: LLM + xTTS + Wav2Lip for personalized greetings.
- Built ~90% of the codebase over 3 months; generated and delivered 300+ personalized videos live at KI Festival Heilbronn.
Tsunami 2025
Freelance · Career Intelligence Agent
- Conversational agent that maps user domains to labor market vulnerability quadrants and generates personalized 3-year upskilling roadmaps.
2023
Salesonic
Freelance · NLP-to-Sales Outreach Pipeline
- Engineered a fully automated NLP-to-sales-outreach pipeline by integrating LLMs with Apollo.io and Salesforge APIs.
- Delivered a functional PoC before the founder's strategic market pivot.
Internal Educational Platform
Frontend Developer · React MVP
- Developed the frontend MVP in ReactJS for an internal educational platform.
- Adopted by 150+ students for project submission and progress tracking.
// research
Bitplane-ANN
8-bit progressive vector search storing embeddings as compact codes instead of float32.
Demonstrates ~4x storage reduction with 99.8-99.9% Recall@10 on 1M-vector text and vision datasets via progressive filtering + reranking.
Agent Security Skill
OWASP-aligned security-review skill for coding agents.
Covers 33 categories across LLM, RAG, MCP, tool, and agentic systems; supports Claude Code, Codex, Cursor, Copilot, and Windsurf.
ATA Protocol
Authorization Type Attestation: IETF Internet-Draft for verifying authorization origin in human- vs AI-authorized communication.
Working PoC includes MCP server and header-level protocol implementation.
CPAR Framework
Cross-provider multi-agent peer review with independent web-searching reviewers.
Outperformed a single-model baseline on 15/15 criteria in blind A/B evaluation.
// hackathons — 3× 1st place
🥇
Future City Hackathon — HHN & 42 Heilbronn
AI agent for automated grid connection requests — end-to-end processing of PV and energy storage infrastructure documents.
November 2025
🥇
Make.com Hackathon — 42 Heilbronn
Automated Next.js internship platform integrating 42 Intra API, Google Sheets, and email workflows.
July 2025
🥇
Wire × Schwarz IT Hackathon
Whisper voice transcription + chat summary AI features. Built the evening before the demo. Directly resulted in an internship offer at Wire.
November 2024
// startup & community
Campus Founders
Judge · Mentor - AI Start 2024 — participant
- AI Start 2025 — pitch finalist
- AI Founders 2025 — founder program
- AI Start April 2026 — technical feasibility judge
- Campus Founders — mentor pool member
CyberValley AI Founders
Pitch Winner Winner of the 2025 pitch competition in the Stuttgart/Tuebingen accelerator ecosystem: 1st place and EUR 10,000 in AWS credits.
AppliedAI Meetup Speaker
Agents Live demo of an autonomous agent architecture in n8n: autonomous loop, skill bank, Python execution sandbox, and Telegram interface in 7 workflow nodes.
42 Heilbronn AI Club
Founder · Organizer Founded and organized the AI Club with 78 members; led hands-on workshops teaching RAG architecture and LLM systems. Built and deployed a Hybrid RAG FAQ agent on the Notion wiki using hierarchical chunking, hypothetical question generation per chunk, and hybrid vector + keyword retrieval to automate new student onboarding queries.
Operational Background
Production 100+ live events per year before AI: concerts, broadcasts, conferences, and high-profile productions where recovery windows were measured in seconds.
Commercial Instinct
Go-to-market Built and toured a commercial cover band; worked in high-ticket real-estate sales; used drone-view marketing and TikTok distribution before it became normal in the local market.
// education
42 Heilbronn — Core Curriculum Certificate
Oct 2023 – Aug 2025 Tuition-free, peer-evaluated software engineering program in Heilbronn, Germany. Completed 6 progressive ranks with 100% pass rate on all exams and 1,800+ hours of project-based C/C++ systems programming.
- Key projects: ft_transcendence — led 5-person team building a real-time multiplayer 3D web platform with WebSockets, Fastify, and React; designed and implemented the AI opponent.
- Systems projects: ft_irc — led 3-person team implementing a full IRC server in C++ from scratch (175h); miniRT — ray tracer rendering 3D scenes in C (280h); minishell — mini bash with signals, pipes, and env variables in C (210h).
- Skills developed: imperative and OOP programming, algorithms, system programming, networking, Linux/sysadmin, and full-stack web.
- Partner engineering visits: Audi, Mercedes-AMG, MHP, T-Systems, Vector, XL2, Schwarz Digits.
- Industry workshops at 42: Aleph Alpha on LLM fine-tuning, AWS, EY on code licensing in the AI era, Schwarz IT on computer-vision detection.
TUM Campus Heilbronn — Foundations and Application of Generative AI (CITHN2014)
Mar – Jul 2024 Bachelor-level course by Prof. Chunyang Chen, Chair of Software Engineering & AI. Covered Transformer architecture, LLMs, prompt engineering, fine-tuning, open-source models such as LLaMA, and GenAI security and ethics. Presented GreeterAI at the course's final pitch session.
Hochschule Heilbronn (HHN) — Autonomous Systems: Deep Learning
Mar – Jul 2024 Master-level course in Automotive Systems Engineering by Prof. Nicolaj Stache. Covered neural networks, CNNs, transfer learning, GANs, Transformers, and reinforcement learning, with practical modules in NumPy and TensorFlow.
SNU by V. Dal, Ukraine — Bachelor of Finance
2007–2011 Recognized by ZAB in Germany as fully equivalent to a German university degree.
// stack
Python TypeScript / JS C / C++ Kotlin RAG · Graph RAG Hybrid RAG Autonomous Agents MCP RAG Evaluation LangGraph LangChain Langfuse OWASP AI Security FAISS · pgvector · HNSW BM25 OCR · Document AI Whisper LiveKit Gradio · Streamlit FastAPI ReactJS · Next.js Azure AI Foundry Docker · Linux GitHub Actions · Azure DevOps