Curriculum Vitae

Machine Learning Engineer

Pitchbook Data (via SPD Technology)

Paris, France

Working as a machine learning engineer through SPD Technology, I contribute to the development of Navigator, an agentic conversational search platform. My role involves implementing complex workflows using LangGraph and tool orchestration to enable sophisticated, multi-step research capabilities for a system that handles tens of thousands of monthly conversations.

• Architected a stateful directed cyclic graph (DCG) to govern a role-separated, multi-node architecture, utilizing conditional edge logic and self-looping semantics to enable iterative research and controlled convergence
• Designed and implemented API-based tools for a diverse tool registry, utilizing Pydantic schema validation and semantic routing to orchestrate retrieval from heterogeneous sources, including PostgreSQL, Search services, and RAG vector stores
• Scale & Concurrency: Optimized the orchestration layer to support hundreds of concurrent agentic loops and parallel requests, ensuring system reliability during high-concurrency traffic bursts
• End-to-End Evaluation Service: Developed a parallel evaluation service consisting of a suite of dozens of evaluation metrics (including faithfulness, relevance, and hallucination detection) to benchmark system performance across large-scale evaluation samples

Machine Learning Engineer (CIFRE PhD Fellow)

Cyclope.ai (VINCI)

Paris, France

Jan 2023 – Aug 2025 · 2 yrs 8 mos

I led the development and deployment of high-performance ML systems for autonomous vehicle safety and large-scale road surveillance. As part of a CIFRE industrial fellowship, I bridged the gap between academic research at the University of Lille and production-ready applications.

Project: Autonomous Vehicle & Intelligence Systems

• Built a high-throughput serving application handling 300M+ POST requests per day for real-time road awareness
• Engineered an end-to-end multi-GPU pipeline with a processing time of just 200ms
• Implemented Bird's Eye View transformations, camera alignment, and multi-camera tracking systems
• Developed an entity attributes service for vehicle speed estimation, lane positioning, and 3D dimension calculations

Project: Road Surveillance Infrastructure

• Boosted YOLO object detection accuracy from 0.91 to 0.96 mAP and increased inference speed by 60% using OpenVino
• Designed technical specs for distributed systems, managing synchronization and communication across multiple physical machines
• Managed a surveillance application processing 100M+ POST requests daily to client servers

Publications:

• "Towards Lightweight Transformer Architecture: an Analysis on Semantic Segmentation" in ACDSA 2024

Machine Learning Research Intern

University of Toronto

Advised by Dr. Brokoslaw Laschowski and Dr. Alex Mihailidis

Remote - Kyiv, Ukraine

I worked within the Intelligent Assistive Technology and Systems Lab (IATSL) under the supervision of Dr. Laschowski and Dr. Mihailidis. My research focused on making lower-limb robotic exoskeletons "smarter" by enabling real-time environment recognition on hardware-constrained devices.

• Advanced Video Classification: Developed high-performance models for stair environment recognition by merging 2D CNN encoders (MobileNet, MobileViT, EfficientNet) with temporal models (LSTMs, Transformers) and exploring 3D CNNs (MoViNet)
• Achieved 98.3% Accuracy and 98.2 F1-score on the StairNet dataset, significantly improving the safety and reliability of human-robot locomotion
• My work directly contributes to improving daily mobility for individuals with lower-limb disabilities, ensuring accurate classification of hundreds of walking environments daily
• Co-authored several papers in leading journals and conferences, including BioMedical Engineering OnLine and IEEE BioRob

Publications:

• "StairNet: Visual Recognition of Stairs for Human-Robot Locomotion" in BioMedical Engineering OnLine
• "Sequential Image Classification of Human-Robot Walking Environments Using Temporal Neural Networks" in IEEE BioRob 2024 & IEEE ICRA 2023, Computer Vision for Wearable Robotics Workshop

Machine Learning Engineer

Infopulse / Tietoevry

Kyiv, Ukraine

I developed and deployed production-grade ML solutions for a major energy sector client, bridging the gap between advanced research and scalable infrastructure.

Project: NOROG (Energy Sector Intelligence)

• Architected ML infrastructure and led the full-cycle implementation of data science initiatives for Offshore Norge, working alongside backend and DevOps teams
• Enhanced search fulfillment from 0.55 to 0.84 (Recall@5) by implementing contrastive learning with CLIP image embeddings and semantic text retrieval
• Fine-tuned a Distil-BERT model for token classification and advanced semantic item filtering, improving search precision by 20% while significantly reducing inference latency
• Deployed models using Triton Inference Server, handling ~30k daily gRPC requests and managing databases (Milvus & PostgreSQL) exceeding 2TB
• Data Lifecycle: Managed end-to-end dataset curation and complex annotations, including recommendation rankings and multi-task text classification

Project: Sales Pitches & Demos

• Engineered a multi-camera person tracking system for sales demos using YOLO v8 and the ByteTrack tracker to showcase real-time monitoring capabilities

Machine Learning Research Intern

Samsung Research

Kyiv, Ukraine

I worked within the Intelligent Security Lab at Samsung Research, focusing on moving state-of-the-art AI from research papers onto actual mobile hardware.

• On-Device Speaker Identification: Deployed the Google TRILL model on mobile hardware, implementing user-specific fine-tuning to enhance authentication accuracy
• Audio Spoofing Detection: Engineered an authentication layer using S-vectors and RawNet models, trained on proprietary in-house datasets to prevent mobile security breaches
• Efficient Video Processing: Integrated the CoViAR (Compressed Video Action Recognition) model to enable high-efficiency video storage and processing directly on mobile devices
• Cross-Platform Development: Utilized TensorFlow Lite, ONNX, and C++ to ensure seamless model integration across mobile ecosystems

Machine Learning Engineer

Anadea

Kyiv, Ukraine

Developed end-to-end ML solutions and Proof-of-Concept (PoC) demos for international clients in e-learning, real estate, and social media. Collaborated with industry leaders, including Zillow, to deliver high-impact data science projects.

Project: Accessible Real Estate (Image Captioning)

• Architected a full-cycle Image Captioning system combining Object Detection with GPT-based text generation to assist users with vision impairments
• Improved BLEU score from 0.39 to 0.48 by iterating on a Meshed-Memory Transformer baseline
• Managed the entire data lifecycle: collection, cleaning, labeling, and versioning (DVC)

Project: Social Media Content Integrity (Duplicate Detection)

• Implemented a Multi-Modal (Vision-Text) system using CLIP to identify duplicate content across high-volume aggregation feeds
• Boosted F1-score from 0.76 to 0.94, leading to significant storage cost savings and improved feed quality
• Collaborated in an agile team of 3 ML engineers to design technical specifications and deployment strategies

Machine Learning Engineer

InDevLab

Kyiv, Ukraine

• Leveraged BERT-based embeddings to build a multi-level matching system for complex item-text data
• Fine-tuned Transformer models to achieve a significant performance jump, increasing the F1-score from 0.62 to 0.92
• Successfully deployed the BERT-powered solution, resulting in a 30% reduction in false matches in the live production environment

Investment Banking Intern

Concorde Capital

Kyiv, Ukraine

• Conducted in-depth market research and competitive analysis across the IT, Agriculture, and Logistics sectors to identify acquisition targets
• Analyzed financial data to support valuation models and investment memorandums
• Prepared pitch decks and presentation materials for senior management and potential clients

Data Analyst Intern

YouScan

Kyiv, Ukraine

• Analyzed and classified large volumes of social media data for multinational enterprise clients using YouScan's monitoring platform
• Generated comprehensive activity reports to track brand sentiment and market trends
• Collaborated with the research team to improve data accuracy and classification logic

Ph.D. in Computer Science

University of Lille 1 Sciences and Technology

Sep 2023 – Aug 2025 (CIFRE Industrial Fellowship with Cyclope.ai)

Conducted advanced research in Deep Learning and Computer Vision for lightweight transformer architectures applied to real-time video anomaly detection. Dropped out after completing two years of the doctoral program to pursue full-time industry leadership in August 2025.

M.Sc. in Applied Mathematics

National University of Kyiv-Mohyla Academy (NaUKMA)

Summa cum laude — 97.45 / 100 GPA

B.Sc. in Applied Mathematics

National University of Kyiv-Mohyla Academy (NaUKMA)

Summa cum laude — 93.78 / 100 GPA