PubMatic is a publisher-focused sell-side platform for an open digital media future. 

Featuring leading omni-channel revenue automation technology for publishers and enterprise-grade programmatic tools for media buyers, PubMatic's publisher-first approach enables advertisers to access premium inventory at scale. 

Processing over one trillion ad impressions per month, PubMatic has created a global infrastructure to drive publisher monetization and control over their ad inventory. 

Since 2006, PubMatic's focus on data and technology innovation has fueled the rise of the programmatic industry as a whole. Headquartered in Redwood City, California, PubMatic operates 13 offices and six data centers worldwide.

We are looking for a strong Machine Learning Intern to join us this summer - a proven 'doer' to develop, implement and extend data-intensive machine learning software for real-time auctioning, ad inventory estimation, and audience segmentations.

You will design and implement core components of our algorithms, as well as model and monetize the large amounts of data that PubMatic generates daily.

Working with our Data Science and Ad Serving teams, you will apply Machine Learning to help get things done.

Responsibilities:

• Development and implementation of data-intensive machine learning software for real-time auctioning, ad inventory estimation, audience segmentations, and other AdTech applications
• Working with data scientists, product managers, and software engineers to develop and support the software for new Machine Learning products
• Ensuring excellence in delivery to internal and external customers

• Actively working towards an MS or PhD degree in a STEM field
• Experience designing and building large-scale ML algorithms and ETL that are well-designed, cleanly coded, well-documented, operationally stable, and timely delivered

And a mix of experience with:

• Python or R, including ML libraries (SKLearn, NumPy, caret, e1071), including CPU/GPU parallelization, matrix algebra, vectorization, linear programming, lambda programming, OOP
• And at least one of the DL frameworks (TensorFlow, PyTorch, Caffe, Theano, Keras, or alike)

With an understanding of:

• Graduate statistics and probability (inference, hypothesis testing, p-value, ANOVA, CLT, LLN, Bayes’ theorem, A/B testing, combinatorics, PDF/CDF, joint/conditional/marginal densities)
• Vector calculus (gradients, Jacobians, partial derivatives and integrals, optimization)
• Linear algebra (eigen values/vectors, inverses, decompositions, orthogonality, multi-linear)
• Time series (ARIMA, GARCH, forecasting, Kalman filter)
• Shallow ML algorithms: regressions, SVM, kMeans, kNN, NB, HMM, PCA, NMF, SVD, XGBoost, decision trees, ensemble methods (random forest)
• Deep NN algorithms: MLP, RNN, LSTM, CNN, GRU
• ML concepts: backprop, hyperparameter tuning (Bayesian optimization, grid/random search), regularization, learning rate, optimization
• Advanced work with SQL or NoSQL, including nested/join/aggregate queries, stored procedures, over partition by, basic stat functions
• Cloud compute engines (AWS, Azure, GCP and alike), ML on clusters of GPUs, SageMaker, Jupyter
• Excellent communication skills, cultural fit and natural curiosity in learning the ML developments and domain expertise

Nice to Haves:

• Prior experience with programmatic advertising / RTB
• Deep reinforcement learning (Bellman equations, MDP, policy optimization, credit assignment, multi-agent)
• Proficiency with Spark (ML Lib, GraphX), Hive or Hadoop
• Scala, Java, C/C++
• A record of STEM publications in top journals or conferences
• And a high rank in Kaggle competitions

PubMatic is proud to be an equal opportunity employer; we don’t just value diversity, we promote and celebrate it.   

We do not discriminate on the basis of race, religion, color, national origin, gender, sexual orientation, age, marital status, veteran status, or disability status.

All your information will be kept confidential according to EEO guidelines.