CX360 ML Engineer

About the project

We're searching for a dynamic Lead Data Scientist focusing on rapid feature and algorithm implementation to spearhead Retail/E-commerce Solution Accelerator for customer lifetime value (CLV) maximization. This role demands a deep understanding of retail/e-commerce dynamics and an agile approach to developing and deploying critical data models.

Today’s AI solutions targeting customer engagement are focused on click-through-rate optimization - raking systems, recommender systems, etc. This helps to optimize engagement at the moment, but this myopic view doesn’t consider long-term goals and customer retention. We are building a new layer for CLV optimization with reinforcement learning on top of existing solutions.

Duration: 6+ months
Stage: solution development from scratch

Areas of Responsibility

• Research of AI solutions and algorithms related to LTV and churn prediction, recommender systems, experimentation, and optimal control

• Hands-on from-scratch implementation of the above-mentioned algorithms in Python

• Communication with the technical leadership and engineering pears

Skills

• Proficiency in Python for quickly implementing machine learning algorithms

• Expertise with experimentation frameworks (SciPy, PyMC3, Spark MLlib) and analytics tools to quickly run robust tests and analyze results. The ability to design and analyze multivariate, non-standard experiments.

  • Example:

    • Leveraged Bayesian hierarchical models in PyMC3 to implement multivariate uplift modeling in Spark to test effectiveness of personalized email campaigns. Quantified incremental response rates.

    • Built out an experimentation framework on SciPy and statsmodels to test the effects of recommendation algorithms, email timing, and coupon targeting on customer conversion rates. Used MANOVA to model the interactions.

    • Developed an revenue optimization engine using contextual bandits algorithms with Spark MLlib. Evaluated performance through A/B testing framework on user randomization.

• Strong optimization algorithms abilities, including reinforcement learning, bandits, and uplift modeling.

  • Reinforcement Learning - Ability to implement temporal difference learning, deep Q-learning, policy gradient methods to optimize long-term rewards. Critical for modeling customer interactions over time.

  • Contextual Bandits - Expertise with bandit algorithms like upper confidence bound, LINUCB, and Thompson sampling to optimize actions based on user context. Key for personalization.

  • Uplift Modeling - Proficiency in techniques like two-model, meta-learner uplift to identify causal impacts of interventions. Crucial for targeting high incremental value customers.

• Data Preprocessing and Engineering: Mastery in transforming complex retail/e-commerce datasets into model-ready formats. Example: Engineered features from raw transaction logs that improved a churn prediction model's accuracy by 20%.

Knowledge

• Deep Understanding of Retail/E-commerce Metrics and Dynamics: Knowing the specifics of retail such as seasonality, purchasing behavior, customer segments, etc.

• Propensity, Uplift, and CLTV Prediction Methodologies: Comprehensive knowledge of these models' theoretical underpinnings and latest trends.

• State-of-the-Art RL Techniques for Optimization: Awareness of current advancements like Deep Reinforcement Learning, Proximal Policy Optimization, etc.

• Advanced Bandit Strategies: Understanding of variations like epsilon-greedy, UCB (Upper Confidence Bound), Thompson sampling, etc.

Experience

• Experience in building AI solutions from research to production 5+ years

• Experience in a fast-moving startup (B2B, A/B/C/D - rounds) or e-fast e-commerce companies environment is a must

Terms & conditions

• Allocation: 0.5+ FTE

• Time zone: preferably Europe

• Candidate’s location: preferably Europe

• Start date: October 2023