commit 287a41e0efa626c2621fa58e593534137cdb50be
parent 51d0bb4f6058e962f09d2c2fc54480c6160d245b
Author: Silas Brack <s174433@student.dtu.dk>
Date: Tue, 27 Feb 2024 22:15:10 +0100
.
Diffstat:
| M | cvs/full.tex | | | 75 | +++++++++++++++++++++++++++++++++++++++++++++++---------------------------- |
1 file changed, 47 insertions(+), 28 deletions(-)
diff --git a/cvs/full.tex b/cvs/full.tex
@@ -12,12 +12,14 @@
$\mid$
\href{https://linkedin.com/in/silasbrack/}{linkedin.com/in/silasbrack/}
$\mid$
- \href{mailto:silasbrack@gmail.com}{silasbrack@gmail.com}
- $\mid$
+ % \href{mailto:silasbrack@gmail.com}{silasbrack@gmail.com}
+ % $\mid$
\href{https://github.com/silasbrack/}{github.com/silasbrack/}
}
-\newsection{Work experience}
+% Data scientist / machine learning engineer with a track record of solving problems with machine learning, from stakeholders to algorithm to production.
+
+\newsection{Work Experience}
% Demonstrate soft skills
% Highlight biggest contributions
@@ -33,13 +35,29 @@
% - Independent
\newtopic
+{Data Scientist (Manager)}
+{Jan 2024 -- \textsl{Present}}
+{Saxo Bank A/S --- Department of Predictive Modelling and Client Research}
+{Copenhagen, Denmark}
+{\begin{accomplishments}
+ \item Promoted to "manager" title (individual contributor branch) due to independence and track record of exceeding expectations.
+ \item Developed a representation learning system for modelling client in-platform behaviour using state-of-the-art transformer models.
+ \end{accomplishments}}
+
+\newtopic
{Data Scientist (Senior Associate)}
-{Apr 2023 -- Present}
+{Apr 2023 -- Dec 2023}
{Saxo Bank A/S --- Department of Predictive Modelling and Client Research}
{Copenhagen, Denmark}
{\begin{accomplishments}
- \item Responsible for user platform behavioural data and stakeholder management for exposing this data to the business for solving business problems.
- \item Responbile for initiating, implementing and productionising representation learning on user platform behavioural data for client modelling.
+ \item Deployed an internal translation application and integrated it into the business to replace ad-hoc human translation workflows, reducing average time spent waiting for translations from XXX to XXX (and saving XXX€ in translation fees).
+ % \item Analysed user behavioral data on the investing platform, discovered ``low-hanging fruit'' triggers which most efficiently identified clients at risk of churn; presented these results to stakeholders.
+ \item Developed metrics for determining the impact of the similarity service and measured the consequent uplift; presented these results to upper management, Chief Sales Officer, etc.
+ % \item Wrote job application for hiring new data engineer.
+ % \item Improved throughput of writing data to Kafka from 130 kB/s to 13 MB/s.
+ \item Re-designed chatbot using open-source embeddings models (...)
+ \item Standardised data science workflows and development environments to incorporate Docker containers for development and consistent linting and formatting (among others), significantly improving onboarding time for the team.
+ \item Commended for repeatedly taking on high-priority tasks during times of excess load, being accountable for demonstrating results, and finishing tasks ahead of schedule---given a raise after three months and, additionally, promoted to data science manager (IC path) after nine months.
\end{accomplishments}}
\newtopic
@@ -47,11 +65,13 @@
{Sep 2021 -- Mar 2023}
{Saxo Bank A/S --- Department of Predictive Modelling and Client Research}
{Copenhagen, Denmark}
-{\begin{accomplishments}
- \item Trained LightGBM models in Python using Spark and DataBricks on data in Kafka, Microsoft Azure Data Lake, and Microsoft SQL Server databases containing hundreds of gigabytes of data.
- \item Upgraded existing LightGBM models to train in parallel using Spark and Synapse ML, significantly improving scalability; scheduled these models to train automatically using Airflow.
- \item Developed a representation learning model for finding similar stocks, resulting in our team winning an internal Hackathon competition for improving client experience. This model was then packaged, containerised in Docker, orchestrated as a microservice in Kurbernetes and exposed via FastAPI/OpenAPI, yielding a response time of under \SI{10}{\milli\second}. As of March 2023, this system has been integrated into official Saxo Bank trading and investing platforms, which is used by more than \SI{750000}{} active clients.% This tool was found to increase customer engagement by XXX\%, thereby improving the Google Quality Score and reducing advertising costs by XXX€ per year.
- \end{accomplishments}}
+{
+ \begin{accomplishments}
+ \item Developed a machine learning model for finding similar stocks, resulting in our team winning an internal Hackathon competition for improving client experience. This model was then packaged, containerised in Docker, orchestrated as a microservice in Kubernetes and exposed via FastAPI/OpenAPI, yielding a response time of under \SI{10}{\milli\second}. As of March 2023, this system has been integrated into the official Saxo Bank trader and investor platforms, which is used by nearly one million active clients.% This tool was found to increase customer engagement by XXX\%, thereby improving the Google Quality Score and reducing advertising costs by XXX€ per year.
+ \item Upgraded existing LightGBM models to train in parallel using Spark and Synapse ML, increasing training dataset from tens of millions of rows to hundreds of millions; scheduled models to train automatically using Airflow.
+ \item Trained LightGBM models in Python using Spark and DataBricks on hundreds of gigabytes of data contained in Kafka, Microsoft Azure Data Lake, and Microsoft SQL Server databases.
+ \end{accomplishments}
+}
\newtopic
{Software Developer (Student Assistant)}
@@ -60,36 +80,35 @@
{Copenhagen, Denmark}
{
\begin{accomplishments}
- \item Automated data engineering tasks and data pipelines with ETL jobs using SSIS and Microsoft SQL Server, being responsible for \SIrange{15}{20}{} jobs.
+ % \item Automated data engineering tasks and data pipelines with ETL jobs using SSIS and Microsoft SQL Server, being responsible for \SIrange{15}{20}{} jobs.
\item Communicated with data cleansing stakeholders to identify and automate the repetitive workflows which led to the most unnecessary time consumption, saving an estimated 1000 man-hours of manual cleansing work. Increased fraction of cleansed data from \SIrange{25}{50}{\percent}.
- % \item Improved technical strategy planning with teams of manual data cleansers during work trip abroad. % More specific, demonstrate soft skills
- \item Created and maintained over \SI{20}{} tools for data manipulation and visualisation using C\#.
+ % \item Created and maintained over \SI{20}{} tools for data manipulation and visualisation using C\#.
\end{accomplishments}
}
\newsection{Projects}
\newtopic
-{Large-Scale Laplace Approximations with Less Hessian Hacks}
-{Sep 2022 -- Mar 2023}
-{In collaboration with S{\o}ren Hauberg, Hrittik Roy, and others}
+{Marginal Likelihood Training of Linearized Laplace Approximations Without Hessian Reductions}
+{Sep 2022 -- May 2023}
+{Supervised by S{\o}ren Hauberg}
{Copenhagen, Denmark}
{\begin{accomplishments}
\item Developed a novel method for computing the Laplace approximation using only Jacobian-vector products in JAX so as to sample from a multivariate normal distribution and compute the log-determinant of the Laplace covariance without explicitly instantiating this covariance.
\item Developed a robust method for evaluating the quality of the approximate sampling algorithm under different configurations.
% \item Formulated multiple preconditioning strategies for the approximate sampling algorithm and demonstrated their theoretical properties.
- % \item Published a paper to NeurIPS 2023~{[2]}.
+ % \item Published a paper to NeurIPS 2023~{[2]}. arXiv: \href{https://arxiv.org/abs/2302.01332}{\texttt{2302.01332}}. Github: \href{https://github.com/eugene/laplace-large-scale}{\texttt{eugene/laplace-large-scale}}.
\end{accomplishments}}
\newtopic
{Bayesian Metric Learning for Uncertainty Quantification in Image Retrieval}
-{Feb 2022 -- May 2022}
-{In collaboration with Frederik Warburg, Marco Miani, and S{\o}ren Hauberg}
+{Feb 2022 -- May 2023}
+{Supervised by S{\o}ren Hauberg}
{Copenhagen, Denmark}
{\begin{accomplishments}
\item Implemented Hessian calculations on contrastive loss for use for Laplace approximation in Bayesian neural networks.
\item Evaluated uncertainty estimates of trained models using the Laplace approximation for metric learning (LAM).
- \item Published a paper to ICML 2023~{[1]}. arXiv: \href{https://arxiv.org/abs/2302.01332}{\texttt{2302.01332}}. Github: \href{https://github.com/FrederikWarburg/bayesian-metric-learning}{\texttt{FrederikWarburg/bayesian-metric-learning}}.
+ \item Published a paper to NeurIPS 2023~{[1]}. arXiv: \href{https://arxiv.org/abs/2302.01332}{\texttt{2302.01332}}. Github: \href{https://github.com/FrederikWarburg/bayesian-metric-learning}{\texttt{FrederikWarburg/bayesian-metric-learning}}.
\end{accomplishments}}
\newsection{Education}
@@ -101,8 +120,8 @@
{Copenhagen, Denmark}
{\begin{accomplishments}
% \item \textsc{Gpa}: 8.7 / 12.
- \item Specialisation in \emph{machine learning and signal processing}
- \item Thesis: ``Effortless Bayesian Deep Learning: Tapping Into the Potential of Modern Optimizers'', supervised by S{\o}ren Hauberg
+ \item Specialisation in \textsl{machine learning and signal processing}
+ \item Thesis: ``Effortless Bayesian Deep Learning: Tapping Into the Potential of Modern Optimizers,'' supervised by S{\o}ren Hauberg
\end{accomplishments}}
\newtopic
@@ -113,11 +132,11 @@
{\small\begin{accomplishments}
% \item \textsc{Gpa}: 9.8 / 12.
\item Completed 35 ECTS of graduate-level courses in machine learning, statistics, mathematics, physics and econometrics.
- \item Specialisation in \emph{cyber materials}
- \item Thesis: ``Membrane Capacitive Desalination'', supervised by Johan Hjelm
+ \item Specialisation in \textsl{cyber materials}
+ \item Thesis: ``Membrane Capacitive Desalination,'' supervised by Johan Hjelm
\end{accomplishments}}
-\newsection{Skills} % Should have all of these in your experience / projects
+\newsection{Skills}
\begin{tabular}{lp{15.5cm}}
\textbf{Languages} & Python, R, SQL, C\# \\
\textbf{Technologies} & Spark, Docker, Kubernetes, Airflow, git \\
@@ -134,8 +153,8 @@
\newsection{Publications}
\begin{accomplishments}
- \item[{[1]}] Warburg et al. `Bayesian Metric Learning for Uncertainty Quantification in Image Retrieval.' In: \textit{40\textsuperscript{th} International Conference on Machine Learning}. 2023.
- % \item[{[2]}] Brack et al. `Large-Scale Laplace Approximations with Less Hessian Hacks' In: \textit{Advances in Neural Information Processing Systems}. 2023.
+ \item[{[1]}] Warburg, Miani, Brack, and Hauberg. `Bayesian Metric Learning for Uncertainty Quantification in Image Retrieval' In: \textsl{Advances in Neural Information Processing Systems}. 2023.
+ % \item[{[2]}] Roy et al. `Marginal likelihood training of linearized Laplace approximations without Hessian reductions' In: \textsl{Advances in Neural Information Processing Systems}. 2023.
\end{accomplishments}
\end{document}
