// portfolio / v1 / 2026
Maksim Khovrov/Novi Sad, Serbia/GMT+2

AI Automation EngineerI build AI agents that run in production.

4.5 years in development and automation, 2+ focused on LLM systems and AI agents. Open to remote.

Try the demosDownload CVContact

Don't take my word for it — try four live demos below.

01numbersanonymized / production figures
voice robots
30+

launched into production

medical / retail / recruitment / finance / municipal

rag accuracy
70 -> 98%

3 pipelines / Qdrant + Zep

deals supported
$400K+

15+ technical audits + commercial proposals

team scaled
3 -> 10+

during automation team restructuring

sectors shipped->
medicalretailrecruitmentfinancemunicipal infrastructureb2c support
02interactive demos
module 01 / active

Personal Assistant + Mnemosyne

Grounded chat: a profile-pinned RAG broker decides what's public before the model ever sees it.

1 / 4
personal_ai · mnemosyneready
model
assistantmatch 100%
Ask anything about my work. I answer only from public sources.
PublicSystem
integrations/ click to expand
scope

This assistant can't leak private data — not by choice, but by design. Its retrieval token is pinned to a public-only profile, so private context never reaches the model in the first place.

try this
  • Ask something out of scope.
  • Ask a stack-specific question.
03experience / operating model
role

Project Manager / Automation Engineer

MCN Telecom / Business Process Automation department
2020 - 2024 / 4 yrs 5 mo
key wins
  • 01Launched 30+ voice robots into production across medical, retail, recruitment, finance and municipal sectors.
  • 02Designed and deployed 3 RAG pipelines with Qdrant and Zep, improving reported response accuracy from 70% to 98%.
  • 03Stood up an internal voice AI with RAG for company B2C customer interactions.
  • 04Conducted 15+ technical audits and supported commercial proposals that closed $400K+ in deals.
  • 05Core developer through restructuring: onboarded specialists and scaled the team from 3 to 10+.
responsibilities
/Integration architecture with Bitrix24, amoCRM, 1C and IP telephony.
/Custom business logic in n8n with JavaScript and Python functions.
/RAG pipeline ownership: dataset prep, vector store config, retriever tuning.
/Prompt engineering and scenario design for higher LLM accuracy.
/Alerting and error handling for external APIs.
/Mentoring, code review, internal knowledge-base maintenance.
how i operate

Operating loop

01
requirements + decomposition
Translate ambiguous asks into specs and testable cases.
02
integration architecture
Sketch data flow, external systems and failure modes before coding.
03
build + ship
Use JS, Python and n8n with version control, review and runbooks.
04
evaluate + tune
Use eval sets, prompt iteration, retriever tuning and latency budgets.
05
monitor + handoff
Add alerting, Swagger docs and knowledge transfer to the next owner.
education
MIREA — Russian Technological University
Computer Science and Software Engineering
04technical stack
languages
JavaScriptTypeScriptPythonSQL
ai / data
LangChainRAGQdrantZepPineconeOpenAI APIPrompt EngineeringFine-tuning
automation
n8nRESTSwaggerDockerCI/CDGitGitHubLinuxJupyter
data / crm
PostgreSQLJSONBitrix24amoCRM1CVoIPJiraNotion
05contact
contact

Hire a reliable AI engineer.

Open to remote AI development, automation, or technical product management. Based in Novi Sad (GMT+2), comfortable with EU and US-East overlap.

Privacy note: demo interactions may be logged for diagnostics. Raw IP addresses are not stored.

emailmkhovrov01@gmail.comopenlinkedinin/maksim-khovrov-113633293opengithubgithub.com/mkhovrov01open