Job Details
Profile (must have skills and experience) we are looking for Statistics - Always makes data-driven decisions using tools from statistics, such as: populations and sampling, normal distribution and central limit theorem, mean, median, mode, variance, standard deviation, covariance, correlation, p-value, expected value, conditional probability and Bayes's theorem Machine Learning Solid grasp of attention mechanism, transformers, convolutions, optimisers, loss functions, LSTMs, forget gates, activation functions Can implement all of these from scratch in pytorch, tensorflow or numpy Comfortable defining own model architectures, custom layers and loss functions Modelling Comfortable with using all the major ML frameworks (pytorch, tensorflow, sklearn, etc) and NLP models (not essential) Able to pick the right library and framework for the job.
Capable of turning research and papers into operational execution and functionality delivery Programming skills: Equipped with programming skills to self-sufficiently manipulate, analyse, synthesise, and visualise data in eg matplotlib, plotly, bokeh, pandas Experienced with testing frameworks (e.
gPyTest) Basic knowledge of SQL Advanced knowledge of design and usage of relational databases (SQL, PostgreSQL) is an added plus Experienced with Docker containers in the context of training and deploying ML models Business Requirement Translation - Able to take ambiguous business requirements, work with stakeholders to clarify them, and propose specific approaches to meet them.
Comes prepared with a view on which option to pursue, incorporating lenses including utility, delivery time, complexity, maintainability, and ethics Research - Proactively identifies shortcomings and problems with current state of the art as applied to Kalido's needs and proposes novel approaches of solving these issues that are not tackled by the current literature Note: Candidates with a background in mathematics, specifically with a good theoretical understanding of concepts such as: vectors, matrices, transposes, inverse, determinant, trace, eigen vector/values, dot product are generally more successful in the interview process