Dmitrii Kovrizhnukh

AI Researcher

Tbilisi, Georgia5 yrs 6 mos experience

Most Likely To SwitchAI Enabled

Key Highlights

Developed innovative RAG systems enhancing ETL processes.
Built scalable NLP services with high accuracy.
Improved payment UX, reducing latency significantly.

Stackforce AI infers this person is a Data Scientist specializing in SaaS and Fintech industries.

Contact

Skills

Core Skills

Machine LearningNatural Language Processing (nlp)Data ScienceData Engineering

Other Skills

Pandas (Software)PostgreSQLdockerElasticSearchDocker ProductsPythonSQLArtificial Intelligence (AI)Front-End DevelopmentSystem DeploymentSite Reliability EngineeringDjangoflash attentionPEFTGenerative Pre-Training

About

Data Scientist with 5 years of commercial experience. Worked in teams of 3–25 people, both onsite and remotely. Experienced in building ML models from scratch: from raw data collection and task tracking to final metrics, reporting, and business forecasting. Strongest in classical ML and NLP, also worked with clustering and CV. Taught ML and analytics at MIPT for 2 years, currently teaching ML at the Department of Machine Learning and Digital Humanities. Active IT community participant and conference speaker. Also run a YouTube channel for aspiring data scientists. Taught "ML and Data Analysis" in "Code of the Future" program for school students (grades 7–10). Contact: somebody84945@gmail.com | Telegram: @MitrofanDS

Experience

5 yrs 6 mos

Total Experience

1 yr 10 mos

Average Tenure

2 yrs

Current Experience

Kanda software

Senior Data Science | NLP Researcher (Acceldata)

May 2024 – Present · 2 yrs · Tbilisi (remote) · Remote

Developed a RAG system for analysts, optimizing the ETL process and increasing coverage by 32.3% per month.
Conducted a successful pilot project of the solution in a client environment, achieving 68.2% customer satisfaction.
Implemented attribute matching functionality, allowing us to process over 15,000 attributes in a week instead of a month, speeding up the overall process by 1.9 times.
Created logical data model functionality, accelerating data annotation by 7.65 times with 81% accuracy on routine cases.
Increased service speed by 43% through caching and query restructuring, which also accelerated work and advanced deadlines.
Implemented user hints for analysts to solve complex problems.
All of the above cases allowed us to shift the migration timeline by 4.1 months.
Translated customer business goals into a full ML specification and delivery plan.
Using retrospectives and dividing the analyst's tasks into blocks, I created an adhesion system that breaks down user requests into these blocks, enabling me to implement two new features without developing new functionality.
Designed the overall ML architecture and supervised model development, testing, and deployment.
Coordinated team work and ensured timely pilot delivery under limited resources.
Successfully launched the pilot on client infrastructure, validating feasibility and achieving targeted performance metrics.

Pandas (Software)PostgreSQLMachine LearningNatural Language Processing (NLP)

Dla piper

Data Science | NLP Researcher

Sep 2022 – May 2024 · 1 yr 8 mos · Грузия · Remote

NLP: built a Q&A; service with 15M documents using ElasticSearch. Deployed inference via Mojo
and Triton, choosing Triton for better latency. UI made with Gradio. Docker-based service with
12–14 sec response time.
Data Analyst: ran A/B tests to keep document search times below threshold, reducing hotline load.
Some solutions increased revenue by 5–7% without affecting other metrics.
Automated ETL with Airflow, saving several hours of processing daily.
Built XGB-based search with 94% accuracy and ~0.92 recall, outperforming CatBoost model.
Created 10+ ML services from scratch with Docker and backend integration; migrated some to
Kubernetes.
Resume updated 10 September 2025 at 11:05
Automated reporting with Airflow + TensorBoard for management analytics.
CV project: waste detection system with Yolov8 + Roboflow annotation.
Skills: Python · Natural Language Processing (NLP) · docker · Machine Learning · Data Science

dockerPandas (Software)Natural Language Processing (NLP)Machine Learning

Acronis

Data Scientist

Sep 2020 – Jul 2022 · 1 yr 10 mos · Georgia · Remote

Worked in the payment experience squad, improved the UX of order confirmation
Did data-pipelines with basic metrics and reports.
Made tlsh hash clusterization, that improved algorithm of hash finding at 72%
Based on the history of the servers, proposed a new scheme of operation, based on which the
latency was reduced by 30% by random forest model.
On the basis of client and server parameters, we compiled a universal configuration that reduced
backups failures by 16%
Technologies: C++, Python, Data Analysis, SQL, PyTorch, Pandas, Numpy, Big Data, Airflow,
Tensorboard, Kubernetes, Git, Grpc Based on the history of the servers, proposed a new scheme of
operation, based on which the latency was reduced by 30%
On the basis of client and server parameters, we compiled a universal configuration that reduced
backups failures by 16%
Skills: Clusterization · Python · Data Science · Data Engineering · Big Data