Артем

Artem Hordiienko
[открыть контакты](см. выше в блоке «контактная информация»)
78347b175/
AI Engineer Kyiv, Ukraine

EXPERIENCE

Senior AI Engineer 06/2025 - 12/2025
pronomy.io
1) Designed and implemented a full LogicRAG inference architecture (https://arxiv.org/abs/2503.12663)
Built a multi-step reasoning engine that decomposes user queries into subproblems, infers dependency graphs, runs
topological execution, retrieves context via a vector store, and synthesizes accurate, policy-aligned responses.
2) Developed RAG pipeline for customer-support use cases
Designed contextual chunking with automatic overlap optimization for support documentation.
Built a vector store backend with metadata indexing and hybrid search.
Integrated a dataset normalization pipeline for multilingual text.
3) Built an intelligent template-selection system
Constructed a knowledge graph of refund/retention rules.
Implemented automatic extraction of:
refund eligibility,
region-specific policies,
required Zendesk templates,
escalation triggers.
4) Added automated escalation detection using LLM
Built a dedicated module that evaluates conversation history + threat keywords + topic patterns.
Integrated secondary LLM calls with JSON parsing, fallback models, and robust error handling.
Fully asynchronous + thread-safe execution with retries and validation.
5) Reliability, monitoring & error resilience
Added structured logging, trace IDs, error routing, and retry logic across all LLM calls.
Implemented safeguards for:
JSON schema corruption,
assistant hallucinations,
missing templates,
fallback models.
Prepared architecture for Prometheus/Grafana metrics (LLM latency, RAG hit-rate, escalation %).

Senior Python Developer 05/2025 - 09/2025
NDA
Key Responsibilities & Contributions:
Developed and deployed production-grade APIs for body and garment feature extraction, outfit recommendation, and real-
time inference (<2 s).
Implemented and fine-tuned computer vision and machine learning models for 3D body scanning, segmentation, and
clothing fit estimation.
Built robust 3D object generation and manipulation pipelines (e.g., mesh optimization, texture mapping, and size
adaptation) to enable virtual try-on experiences.
Designed modular backend architecture with Python (FastAPI/Flask) ensuring scalability, clean code, and GDPR-compliant
data processing.
Collaborated directly with the CTO and product team to define the technical roadmap, data flow, and integration strategy

1/3
 with pilot retailers.
Optimized model inference using GPU acceleration and batch-based serving pipelines for real-time user interaction.
Created technical documentation and established reproducible pipelines for future 3D/AR personalization extensions.
Contributed to recommendation system prototyping, leveraging embeddings and similarity search for personalized outfit
suggestions.

AI Engineer 11/2023 - 04/2025
AI Docs
LCM architecture for explainability: designed a concept-bottleneck pipeline (concept extractor attention/intermediate
heads classifier/generative heads) so decisions pass through human-readable “concepts” (e.g., qualification clause,
delivery terms, payment terms, document vs. condition), improving auditability and reviewer trust.
Domain concept inventory: curated a reusable inventory of procurement concepts (phrase lexicons, modal verbs, passive
markers, compliance flags) and aligned them with model concepts; enabled transparent rationales in UI and model cards.
Requirement intelligence: shipped document-vs-condition classification, duplicate detection, clause/requirement tagging,
and gap analysis against public-procurement rules to prevent disqualification.
Tender ranking & triage: combined LCM concept scores with rule features (payment terms, product fit, qualification
thresholds, expected margin) to prioritize high-ROI opportunities.
Quality & safety governance: built an evaluation harness (accuracy, latency, cost) and safety tests (prompt-abuse, data-
exfiltration); added calibration checks on concept heads to reduce false positives in compliance-critical flows.

Computer Vision Engineer 05/2024 - 09/2024
Freelance
Launched a “find product by photo” feature: users snap/upload a picture and get the exact item. Result: search product
page conversion up 30%.
Reduced wrong matches by 45%, lowering complaints and returns.
Cut response time to < 5 second, making search noticeably faster and easier to use.
Scaled the system to support 100K–1M images while keeping quality and stability.
Found and merged duplicate listings across sellers, cleaning up the catalog and easing moderation.
Built clear dashboards and KPIs so stakeholders can track conversion, speed, and errors in real time.

Senior Software Engineer 01/2019 - 03/2023
ArbitrageUp
KPIMeter is ML model for analyzing work performance of our marketing team members (private develpment)
Responsibilities & Achievements
Collaborating with the marketing team to understand the requirements and performance metrics to be analyzed.
Designing and developing AI models using PyTorch to analyze the work performance of marketing team members.
Integrating the AI model with MongoDB to log and store the analysis results.
Deploying the AI model on AWS to make it accessible to the marketing team.
Setting up CircleCI for automated testing and continuous integration.
Monitoring the performance of the AI model and making improvements based on feedback and data analysis.
Contributed to sprint planning by defining technical requirements and estimating task complexity, ensuring on-time delivery
of 90% of sprint goals
Technologies used:
Tensorflow
Scikit‑learn
Polaris (Pandas analog)
Mongo DB (for logging results)
AWS (for deploying AI model)
CircleCI (for automating testing)

2/3
E D U C AT I O N

Computer Engineering 09/2019 - 07/2022
Kyiv's Polytechnical University - Bachelor's Degree

C E RT I F I C AT E S

Machine Learning

SKILLS

Python Advanced SQL (Postgres, MS SQL,
Advanced
MySQL)
Scikit-learn Advanced
Polaris Advanced
Tensorflow Intermediate
Pandas Advanced
MongoDB Advanced
Numpy Advanced
Pydantic Intermediate
CI/CD Intermediate
AWS Intermediate
Docker Intermediate

LANGUAGES

English Fluent Spanish Basic
Ukrainian Native Russian Fluent
Swedish Basic

3/3