Senior Applied Scientist

Zepto · Bengaluru

  • Experience4–7 yrs
  • SalaryNot disclosed
  • Work modeonsite
  • Levelmid
  • Posted2 Sept 2026
Apply for this job

About Zepto

Zepto is hiring in Bengaluru in ecommerce retail. This role looks for around 4+ years of experience.

Skills

  • Machine Learning
  • Time Series Forecasting
  • PyTorch
  • Python
  • SQL
  • Pandas
  • Polars
  • scikit-learn
  • statsmodels
  • GluonTS
  • sktime
  • Prophet
  • AutoGluon
  • AWS
  • GCP
  • Azure
  • Docker
  • Kubernetes
  • MLOps
  • GPU Computing
  • Data Structures
  • Algorithms
  • System Design
  • Feature Engineering

The role

An applied scientist at an e-commerce marketplace develops machine learning solutions for demand forecasting and predictive modeling, applying time series forecasting and PyTorch to large-scale demand data, and builds MLOps systems for production. Python and GPU Computing support scalable model training and inference.

Full job description

Requirements:

Experience working with machine learning algorithms and technologies, with hands-on production deployment experience.

Experience working on time series problems, implementing existing methods, and developing new solutions (statistical and probabilistic models; deep learning CNN, LSTM, and transformer architectures).

PyTorch-based, GPU-accelerated training of time series models, including experience optimizing training pipelines for large-scale, high-cardinality demand data.

Deep, hands-on understanding of statistical forecasting methods (ARIMA/SARIMA, exponential smoothing, state-space, and Bayesian structural models) and when to apply them versus ML/DL approaches.

Design, develop, and maintain scalable machine learning systems for time series forecasting and general predictive modeling (on both CPU and GPU machines).

Strong proficiency in Python and the ability to write production-level code.

Critical thinking, strong technical knowledge in data structures, algorithms, and system design.

Key Technical Skills:

Languages/Libs: Python (production-grade); working knowledge of SQL, Pandas, and Polars.

ML/DL Frameworks: PyTorch, scikit-learn, and time series Python libraries (e. g., statsmodels, GluonTS, sktime, Prophet, and Chronos-2 AutoGluon).

Cloud & Infrastructure: AWS, GCP, or Azure; containerization and orchestration (Docker, Kubernetes) for training and inference workloads.

MLOps: Model monitoring, alerting, experiment tracking, and CI/CD pipelines for ML systems in production.

GPU Computing: Experience with GPU-based distributed training and inference for deep learning models.

Data Tools: Experience working with large-scale structured/unstructured data pipelines and feature engineering.

Preferred Skills

Working knowledge of the latest research directions in time series foundation models DeepAR, Chronos-2 and related pretrained/foundational forecasting architectures and the ability to evaluate and adapt them for production use.

Experience building deep learning-based time series models that account for promotional and event effects and supporting inventory optimization simulation setups.

Experience with demand pricing algorithms, including dynamic and elasticity-based pricing approaches.

Exposure to reinforcement learning (RL)-based algorithms for pricing and inventory optimization.

Ability to do exploratory data analysis, read research papers and state-of-the-art models in literature, and implement them.

Potential to innovate novel machine learning methods at industry standards, with preferred publications in top-tier conferences (e. g., NeurIPS, ICML, ICLR, KDD, AAAI).

Exposure to natural language processing and computer vision algorithms is a bonus.

Knowledge of data governance and ethical AI principles.