• File

Nikita

AI-інженер

Considering positions:
AI-інженер, Fullstack-програміст, Python-програміст, Fullstack-розробник 
Age:
22 years
City of residence:
Kyiv
Ready to work:
Remote

Contact information

The job seeker has provided: Phone numberEmailLinkedInMessenger

Name, contacts and photo are only available to registered employers. To access the candidates' personal information, log in as an employer or sign up.

Uploaded file

Quick view version

This resume is posted as a file. The quick view option may be worse than the original resume.

NIKITA LINCHAK
Kyiv, Ukraine | Remote | [open contact info](look above in the "contact info" section) | [open contact info](look above in the "contact info" section) | LinkedIn: nikos3rd | GitHub: TrebaDripat | Telegram: @nikos3rd
Senior Applied AI Engineer | LLM Systems
PROFESSIONAL SUMMARY
Applied AI engineer with 4+ years of commercial experience shipping LLM systems from design to production. Built a LangGraph/FastAPI lead-
research engine that processed 2,000 leads in one production run and two human-in-the-loop outreach systems used by sales managers. Focus
areas: grounded generation, prompt-injection defense, eval-based regression checks, and LLM cost control.
PROFESSIONAL EXPERIENCE
Confidential (B2B sales-automation & HR software company) - AI Solutions Specialist (AI-First Developer)
Jan 2026 - Oct 2026 | 10 mos | Remote
B2B sales-automation and HR software company. Owned AI delivery end-to-end, working directly with the CEO and CTO.
Replaced an estimated 250+ hours of manual SDR research per 2,000-lead run by turning an n8n prototype into a LangGraph/FastAPI service
(Apollo discovery, Perplexity verification, ICP scoring).
Grounded drafts in approved evidence and templates while isolating untrusted content, validated on a 107-lead replay set (82 grounded drafts,
25 justified skips). Checked regressions with golden datasets and LLM-as-a-judge.
Made runs resumable after crashes with PostgreSQL checkpoints; idempotency keys kept retries from repeating paid enrichment calls.
Shipped two Closely (LinkedIn outreach platform) workflows for sales managers: Telegram/web approval before each send, and manager
escalation for pricing, negative replies, and AI disclosure (EU AI Act Art. 50).
Reverse-engineered Closely's private API into a typed client and FastMCP server; deduplication removed ~17% of leads before campaign
import (in validation).
Delivered modules for HR Platform (internal) (TypeScript/Node.js/Express/MongoDB), including a Gemini timesheet assistant. The modules are
in production use at the company.
Maya B LLC (Delikeys) - AI Automation & Data Integrations Engineer
Jan 2025 - Dec 2025 | 1 yr | Remote
E-commerce operations across Mercado Libre and Amazon in Latin American and US markets.
Cut LLM calls ~77% on a multi-thousand-SKU Mercado Libre/Amazon catalog by filtering records before model calls (314 of 407 skipped in a
validation run).
Rewrote product titles in production with Gemini 2.5 Flash on Vertex AI. Validated the 60-character limit and preserved key attributes.
Consolidated three marketplace workflows into one monitored pipeline with isolated sub-workflows and Slack alerts. Cached unchanged titles
by SKU to avoid repeat generation.
UkrLiftService (ULS) - Systems Administrator & AI Automation Engineer
Jul 2022 - Dec 2024 | 2 yrs 6 mos | Kyiv, Ukraine
Sole AI specialist at an elevator manufacturer and service company.
Built hybrid RAG over technical and regulatory documents (BM25 + Sentence-BERT, RRF, cross-encoder reranking) so engineers could find
specs and exact part numbers without searching folders by hand.
Made lead scores auditable in a FastAPI/PostgreSQL CRM by scoring asynchronously with an LLM and storing a structured rationale for each
score.
Built a CV-screening agent that ranks candidates against vacancies and gives a written reason for each match.
KEY PROJECTS
Hybrid Product Search: PostgreSQL FTS, E5/pgvector and BGE reranking. Evaluated with nDCG, MRR and Recall@k.
AmLaw AI Visibility Index: Python pipeline compared citations across three LLMs (611 answers analyzed) with per-run cost limits.
CORE COMPETENCIES
LLM frameworks: LangGraph, LangChain, FastMCP, Model Context Protocol (MCP), Claude Agent SDK.
Models & APIs: Claude, OpenAI, Gemini on Vertex AI, Perplexity, OpenRouter, Ollama.
Agent patterns: Tool calling, Pydantic structured outputs, human approval, prompt-injection isolation.
Retrieval (RAG): BM25, Sentence-BERT, PostgreSQL FTS, pgvector HNSW, E5, RRF, BGE cross-encoder reranking.
Evaluation: Golden datasets, LLM-as-a-judge, nDCG, MRR, Recall@k, abstention.
Backend & delivery: Python, FastAPI, TypeScript, Docker, GitHub Actions, AWS EC2, pytest/respx.
Data: PostgreSQL, Qdrant, Pinecone, Supabase, MongoDB, BigQuery.
LLMOps: Run logs, token/cost tracking, retries and provider fallbacks.
EDUCATION
MSc in Computer Science, Kyiv National University of Technologies and Design (KNUTD), in progress.
BSc in Computer Science, KNUTD. Hybrid information retrieval thesis: “Development of an Information Retrieval System to Enhance E-
Commerce Product Search Efficiency” (89/100).
Research paper/presentation: “Risk Classification of AI Systems Under the EU AI Act”.
LANGUAGES
English (B2) | Ukrainian (native) | Russian (native)

More resumes of this candidate

Similar candidates

All similar candidates

Candidates at categories

Candidates by city


Compare your requirements and salary with other companies' jobs: