Building Soda to monitor from data to decision

I love talking about: machine learning, data quality, decision making, post deployment data science, performance estimation, and bunch of other stuff. 

Always open to have a chat 🙂

Hakim Elakhrass

Wrike

A leading global work management platform (SaaS) used by over 20,000 companies and 2M users worldwide.

Achievements:
1) Owned the entire lifecycle of data pipeline and model development, deployment, automation (Airflow), monitoring, and post-deployment analysis and improvement for several machine learning models, such as:
- Models to predict propensity to churn, downgrades, and upgrades across multiple prediction horizons. These models provided well-calibrated outcomes, contributing to a 11% MAPE in expected loss forecasting from churn and downgrades over the next six months. The model significantly enhanced churn prevention campaigns, reducing churn by 20% compared to the previous solution.. 
- Implemented a Multitouch Attribution model, driving a 59% lift in attributed Expand ARR and unlocking an additional $2.4 M in pipeline from early‑stage and digital touchpoints.
- Recommendation system to improve product adoption using Causal Inference techniques. This model boosted retention by 13% and increased the upgrade rate by 6%.
2) Enhanced the post-deployment model monitoring system by implementing advanced PSI monitoring. Developed visualisation tools and integrated a Slack notification system. These improvements significantly reduced the identification time of data issues to within 1 day of occurrence.
3) Implemented multiple enhancements to the leads LTV prediction model. These improvements resulted in a 17% increase in revenue from high-quality leads.
4) Integrated Causal Inference methodologies across various processes, such as product analysis, recommendation algorithms and experiments evaluation, improving quality and reliability of data-driven decision-making.

Senior Data Scientist

Rubbles

Leading Data Science consultancy

Achievements:
Enhanced the promotions calendar optimiser functionality, including demand (time series) forecasting models for dozens of SKUs across multiple time horizons, and a combinatorial optimiser handling thousands of customer-defined constraints. This included refining the existing forecasting models and deploying the entire model pipeline to Azure Databricks. As a result, there was a 4% (absolute) increase in ROI.

Beeline Russia

Russia's 3rd largest telecom operator with ~45M of subscribers

Achievements:
1) Managed the entire lifecycle of a new churn model, including development, deployment, and post-deployment maintenance. Implemented automation for data quality checks and monthly model retraining in production. This model contributed to a 27% reduction in churn rate.
2) Conducted subscriber clustering based on their identified issues, interpreting and systematising the data. This facilitated targeted improvements, resulting in an additional 11% reduction in churn rate.
3) An engine for analysis of feedback, collected by NPS has been built
4) Enhanced the A/B testing platform by implementing an engine that constructs target and control groups balanced across numerous features of various types. Improvements also included sensitivity evaluation using A/A tests, and tools for post-stratification and CUPED application for more accurate experiment analysis.

Data Scientist

Adamas-Yuvelirtorg

Leading Russian jewellery retailer, e-commerce department

Achievements:
1) Multiple recommender systems have been developed using both classical machine learning and deep learning approaches. Most of these systems utilized a combination of descriptive features and visual features extracted through computer vision techniques. Their combined efforts resulted in a 16% increase in revenue.
2) User showcases personalisation system has been developed. It allowed to increase revenue by 7%.
3) A search system for similar jewellery by photo has been developed.

State Atomic Energy Corporation "Rosatom"

- ML-based tool has been developed for comprehensive interpretation of geophysical logging data.
- Thousands of automated Monte Carlo simulations were performed to model the spectra of neutrons and gamma rays detected by geophysical logging devices.
- Computational optimization of the geometry and operating modes was performed for devices used in neutron-gamma density logging.

Data Mining Specialist

- Development and application of new methods of optimization of radiation shielding for small-sized nuclear power units;
- Modeling of the interactions of cosmic rays with structural elements of space nuclear power units;
- Creation of a safe start procedure for space nuclear power unit in orbit without trigger source;
- Computational optimization of proton target for molten salt sub-critical reactor.

Research Programmer

-Studied the design, operation, and decommissioning of nuclear power installations. 
-Participated in innovative projects focused on the development and improvement of small-scale nuclear power installations through numerical simulations and the application of optimization algorithms. 
-Published 7 scientific articles.

Postgraduate Student

Senior Data Scientist focused on maximizing delivered value. 

I ensure that projects drive concrete, actionable decisions, not just models left unused. Beyond applying predictive Machine Learning, Data Science, AI, statistical analysis, causal inference, and decision science, I see my role as evidence-driven problem solving.

I either implement or technically lead the full model lifecycle, starting with root-cause analysis to identify the real problem or the most promising growth opportunity. By combining growing domain expertise, rapid prototyping and close collaboration with other experts, I design practical, high-impact solutions. And then set up monitoring, observability and continuous-improvement processes after the first deployment. 

Throughout development and post-deployment I keep a tight feedback loop with business stakeholders and end users to build transparency, trust and working strategies that produce measurable results.

taplio

The Battle

Hakim Elakhrass

Ilia Ekhlakov

Followers

Posts

Engagements

Likes

Ready to grow your LinkedIn presence?