Keep in touch with meI'm using Intch to connect with new people. Use this link to open chat with me via Intch app
Work Background
ProvectusLead Machine Learning Engineer
Aug. 2021Toronto, Ontario, Canada· Led a dynamic team of 10+ engineers, regularly conducting internal workshops on recent ML advancements and
spearheading iterative system improvements, resulting in a 20% performance boost over the year
· Mentored 4 team members in writing 2 articles and 10 blog posts in Provectus, resulting in a 70% increase in content quality.
· Implemented a search functionality for Provectus utilizing Retriever-Augmented Generation (RAG), enhancing search
relevancy by 25% and streamlining processes for NOC and IT departments
· Devised a heuristic-based detection framework to identify and neutralize malicious scripts in obfuscated JavaScript,
achieving a detection accuracy of 98% on benchmark datasets
· Implemented focal loss and sample re-weighting strategies to counteract class imbalance in large data sets, improving
recall for minority classes by 15%
· Engineered robust ML pipelines leveraging Sagemaker Pipelines, Step Function Pipeline and Wandb for continuous training
and fine-tuning of GPT-3 models, reducing model drift and improving domain specificity
ProvectusSenior Machine Learning Engineer
Aug. 2019 - Aug. 2021Toronto, Ontario, Canada· Led the development of a text extraction and QA system employing OpenAI GPT-3, achieving a 75% accuracy (EM) on
domain-specific data sets
· Optimized data preprocessing pipelines resulting in a 15% increase in overall efficiency
· Led document multi-class classification efforts using transformer architectures, achieving an F1 score of 087 on extensive
datasets
· Developed a multi-label text classification pipeline using Transformers and Pytorch, handling over 100 labels and achieving
a 70% micro-average precision
· Pioneered the transformation of clinical trial descriptions into SQL-like logic using BERT-based models, reducing error rates
by 20% compared to conventional methods
· Designed and executed benchmarking tests for the QA task within the biomedical domain, specifically focusing on RAG
requests across various LLMs
· Increased model training process efficiency by 30% through the development of a scalable Amazon SageMaker ML pipeline
· Curated a comprehensive dataset from reputable biomedical sources and implemented evaluation metrics, including BLEU
score, ROUGE-L, and EM, ensuring a holistic assessment of model performance in real-world scenarios
· Compiled and optimized YOLOv4 models on Nvidia Jetson Xavier and Nano, achieving a real-time frame rate of 30 FPS on
HD video feeds
· Architected an end-to-end video processing pipeline from edge devices (using Gstreamer) to AWS Cloud for real-time
analytics and report generation using AWS Lambda and S3
· Trained the YOLO model, achieving real-time object detection and tracking, with velocity estimation using mapping to GPS
coordinates
· Developed a real-time industrial appliance monitoring system using machine learning for predictive maintenance
· Applied TF-IDF and Word2Vec embeddings along with encoders with advanced statistical techniques for code segmentation,
resulting in a 30% improvement in feature relevance for item matching
ProvectusMachine Learning Engineer
Jan. 2017 - Aug. 2019Kazan, Tatarstan, Russia· Design and Implement the solution, achieving 95% F1 metric in classifying newborn eye abnormalities.
· Designed backend solutions and user-friendly interfaces, showcasing strong software architecture knowledge.
· Collaborated with cross-functional teams, offering technical guidance for project improvement.
· Acted as the tech lead, coordinating with team members to ensure smooth progress and high-quality results.
· Integrated with video services to obtain the audio stream and process it using speech2text algorithms
· Developed a UI for working with a list of recommendation for emails and responses
· Implemented a pipeline for the Multi-label Classification problem of the emails
· Fine-tuned N-gram and early transformer models to provide real-time sales guidance, resulting in a 15% sales uplift.
· Optimized feature extraction process, leading to a 2% increase in model accuracy.
· Implemented algorithms for courier route optimization, resulting in a 10 minutes decrease in delivery times and a 16.6%
increase in on-time deliveries.
· Integrated ML models into backend infrastructure for real-time client estimations.
ProvectusSoftware Engineer
Jan. 2016 - Jan. 2017Kazan, Tatarstan, RussiaLed the engineering team in a startup environment, optimizing a delivery logistics platform.
Optimized performance of the usage of MongoDB
Spearheading the integration with the vast local online market.
FOSS LabsSoftware Engineer
Oct. 2014 - Jan. 2016Kazan· Implemented a website for Rubin Kazan football club and an admin panel for the Galiaskar Kamal Tatar
Academic Theatre, the premier Tatar theater in Kazan.
BARS GROUPSoftware Engineer
Jul. 2014 - Oct. 2014Kazan, Tatarstan, Russia· Developed numerous comprehensive analytic reports utilizing large datasets for a medical reporting application.
· Enhanced efficiency by optimizing resource-intensive SQL scripts, reducing report generation time from 8 hours to 20
minutes.
Intch is a Professional Networking App for the Future of WorkSee more people like Rinat on Intch
300k+ people
130+ countries
AI matching
Get Intch if you'd like to
1021098 people
15
IT Support Manager @ All In Credit Union
15
Frontend Developer
15
Help Desk Specialist @ Huntington Ingalls Industries, Inc.
83961 people
15
IT Support Manager @ All In Credit Union
21
IT Manager | COO @ Refipampa
15
Independent Consultant @ Independent Consultant