update

apointa · apointa · commit 11e2c8d45c47 · 2026-02-15T12:24:59.000+01:00
diff --git a/content/authors/admin/_index.md b/content/authors/admin/_index.md
@@ -16,22 +16,22 @@ status:
 superuser: true
 
 # Role/position/tagline
-role: Ellis PhD student in Machine Learning
+role: Applied Scientist at AWS (Amazon)
 
 # Organizations/Affiliations to show in About widget
 organizations:
-  - name:  Ellis Unit Linz - Institute for Machine Learning, Johannes Kepler University, Linz
-    url: https://ml-jku.github.io/
+  # - name: AWS (Amazon)
+  #  url: https://ml-jku.github.io/
 
 # Short bio (displayed in user profile at end of posts)
-bio: I am a computer physicist in soft matter and fluids at interfaces, currently at the LIPhy in Grenoble (France).
+bio: 
 
 # Interests to show in About widget
 interests:
+  - Foundation Models
+  - Time Series
   - Deep Learning
   - Uncertainty Estimation
-  - Time Series
-  - Neurosymbolic AI
 
 # Education to show in About widget
 education:
@@ -54,12 +54,12 @@ social:
   - icon: envelope
     icon_pack: fas
     link: '/#contact'
-  - icon: twitter
-    icon_pack: fab
-    link: https://twitter.com/AndAuer
-    label: Follow me on Twitter
-    display:
-      header: true
+  #- icon: twitter
+  #  icon_pack: fab
+  #  link: https://twitter.com/AndAuer
+  #  label: Follow me on Twitter
+  #  display:
+  #    header: true
   - icon: linkedin
     icon_pack: fab
     link: https://www.linkedin.com/in/andreas-auer-cs
@@ -72,9 +72,9 @@ social:
   - icon: github
     icon_pack: fab
     link: https://github.com/apointa
-  - icon: gitlab
-    icon_pack: fab
-    link: https://gitlab.com/apointa
+  #- icon: gitlab
+  #  icon_pack: fab
+  #  link: https://gitlab.com/apointa
   # Link to a PDF of your resume/CV.
   # To use: copy your resume to `static/uploads/resume.pdf`, enable `ai` icons in `params.yaml`,
   # and uncomment the lines below.
@@ -91,12 +91,9 @@ highlight_name: true
 
 **About me**
 
-Hi, I am Andreas. I am a **PhD student** in the field of **Machine Learning** at the [Institute for Machine Learning](https://ml-jku.github.io/) at Johannes Kepler University, Linz, Austria, advised by Sepp Hochreiter. I am also part of the [ELLIS PhD Program](https://ellis.eu/phd-postdoc).
-
-In my research I am most interested in **Deep Learning** in the context of **Time Series** or, more broadly, sequential data.
-Currently I am specifically interested in **Foundational Time Series Models**:
+Hi, I am Andreas. I’m an **Applied Scientist at AWS (Amazon)**, dedicated to advancing **foundation models for time series**, with a specific focus on **forecasting**. I work in the team that leads the research and development of [Chronos](https://github.com/amazon-science/chronos-forecasting) and [Autogluon](https://github.com/autogluon/autogluon).
 
-I led the development of **[TiRex](https://arxiv.org/abs/2505.23719)**, a state-of-the-art foundational forecasting model built with xLSTM, and **[COSMIC](https://arxiv.org/abs/2506.03128)**, the first foundational forecasting model that beneficially utilized covariates in a zero-shot setting.
-Further, I am the co-author of **[Chronos-2](https://arxiv.org/abs/2510.15821)** and **[xLSTM](https://arxiv.org/abs/2405.04517)**.
+Prior to that, I completed my **PhD in Machine Learning** at the [Institute for Machine Learning](https://ml-jku.github.io/) at JKU, where I was advised by [Sepp Hochreiter](https://scholar.google.com/citations?user=tvUH3WMAAAAJ) and was part of the [ELLIS PhD Program](https://ellis.eu/phd-postdoc). My PhD research centered on **Deep Learning** in the context of **Time Series** or, more broadly, sequential data — with a specific focus on Foundation Models. Before that, I obtained my BSc and MSc degrees in Computer Science at the Technical University of Vienna and gathered experience as a professional software developer.
 
-Prior to my PhD, I completed my BSc and MSc degrees in Computer Science at the Technical University of Vienna and gathered experience as professional software developer.
+I led the development of **[TiRex](https://arxiv.org/abs/2505.23719)**, a state-of-the-art foundation forecasting model built with xLSTM, and **[COSMIC](https://arxiv.org/abs/2506.03128)**, the first foundation forecasting model that beneficially utilized covariates in a zero-shot setting.
+Further, I am a co-author of **[Chronos-2](https://arxiv.org/abs/2510.15821)** and **[xLSTM](https://arxiv.org/abs/2405.04517)**.
diff --git a/content/home/contact.md b/content/home/contact.md
@@ -9,7 +9,7 @@ headless: true
 weight: 130
 
 title: Contact
-subtitle: Feel free to send me an email or contact me via twitter
+subtitle: Feel free to send me an email or contact me via LinkedIn
 
 content:
   # Automatically link email and phone or display as text?
@@ -18,10 +18,10 @@ content:
   # Contact details (edit or remove options as required)
   email: auer[at]ml.jku.at
   contact_links:
-    - icon: twitter
-      icon_pack: fab
-      name: Twitter
-      link: 'https://twitter.com/AndAuer'
+    #- icon: twitter
+    #  icon_pack: fab
+    #  name: Twitter
+    #  link: 'https://twitter.com/AndAuer'
     - icon: linkedin
       icon_pack: fab
       name: LinkedIn
diff --git a/content/home/experience.md b/content/home/experience.md
@@ -24,38 +24,38 @@ date_format: Jan 2006
 #   Leave `date_end` empty if it's your current employer.
 #   Begin multi-line descriptions with YAML's `|2-` multi-line prefix.
 experience:
-  - title: PhD Student in Machine Learning
+  - title: PhD in Machine Learning
     company: Johannes Kepler University
     company_url: ''
     company_logo: jku
     location: Linz, Austria
     date_start: '2022-05-01'
-    date_end: ''
-    description: 'My research is currently in the field of Uncertainty Estimation for Machine Learning Models, especially for Deep Learning Models and in the context of Time Series.'
+    date_end: '2026-01-10'
+    description: 'Graduated with best grade - Supervisor Sepp Hochreiter - Research on deep learning in the context of time series with a specific focus on foundation models.'
   - title: Master in Computer Science
     company: Technical University of Vienna
     company_url: ''
     company_logo: tu-vienna
     location: Vienna, Austria
     date_start: '2020-10-01'
     date_end: '2022-04-30'
-    description: Graduated with distinction - GPA 3.98 <br> My curricula focused on intelligent information systems (Symbolic & Neural AI, Database Systems)
+    description: Graduated with distinction - GPA 3.98/4 <br> My curricula focused on intelligent information systems (Symbolic & Neural AI, Database Systems)
   - title: Bachelor in Computer Science
     company: Technical University of Vienna
     company_url: ''
     company_logo: tu-vienna
     location: Vienna, Austria
     date_start: '2017-10-01'
     date_end: '2020-09-30'
-    description: Graduated with distinction - GPA 3.90 <br> General CS program with a focus on Software and Information Engineering <br> I additionally participated in the [Bachelor with Honors program](https://informatics.tuwien.ac.at/bachelor-with-honors/) 
+    description: Graduated with distinction - GPA 3.90/4 <br> General CS program with a focus on Software and Information Engineering <br> I additionally participated in the [Bachelor with Honors program](https://informatics.tuwien.ac.at/bachelor-with-honors/) 
   - title: Bachelor in Economics
     company: Vienna University of Economics and Business
     company_url: ''
     company_logo: wu
     location: Vienna, Austria
     date_start: '2017-10-01'
     date_end: ''
-    description: Side passion
+    description: Concurrent studies as side passion
 design:
   columns: '2'
 ---
diff --git a/content/home/experience_2.md b/content/home/experience_2.md
@@ -24,13 +24,21 @@ date_format: Jan 2006
 #   Leave `date_end` empty if it's your current employer.
 #   Begin multi-line descriptions with YAML's `|2-` multi-line prefix.
 experience:
+  - title: Applied Scientist
+    company: Amazon Web Services (AWS)
+    company_url: ''
+    company_logo: aws
+    location: Berlin, Germany
+    date_start: '2026-02-01'
+    date_end: 
+    description: Research on Foundation Time Series Models.
   - title: Pre Doc Researcher
     company: Johannes Kepler University
     company_url: 
     company_logo: jku
     location: Linz, Austria
     date_start: '2022-05-01'
-    date_end: ''
+    date_end: '2026-02-01'
     description: Pre Doc research associate at the Institute for Machine Learning.
   - title: Applied Scientist Intern
     company: Amazon Web Services (AWS)
@@ -39,71 +47,63 @@ experience:
     location: Berlin, Germany
     date_start: '2025-09-01'
     date_end: '2025-12-30'
-    description: Research on Foundational Time Series Models.
+    description: Research on Foundation Time Series Models.
   - title: PhD Researcher
     company: NXAI GmbH
     company_url: ''
     company_logo: nxai
     location: Linz, Austria
     date_start: '2025-04-01'
     date_end: '2025-08-30'
-    description: Research on Foundational Time Series Models.
+    description: Research on xLSTM-based architectures for time series forecasting.
   - title: Applied Scientist Intern
     company: Amazon Web Services (AWS)
     company_url: ''
     company_logo: aws
     location: Berlin, Germany
     date_start: '2024-07-01'
     date_end: '2024-11-30'
-    description: Research on Foundational Time Series Models.
+    description: Research on zero-shot forecasting and covariate handling in Foundation Models.
   - title: Knowledge and Software Engineer / ML Consultant
     company: Onlim GmbH
     company_url: ''
     company_logo: onlim
     location: Vienna, Austria
     date_start: '2020-08-01'
     date_end: '2023-08-31'
-    description: First I worked part-time as a Backend Developer during my master's studies, focusing on enhancing the dialog component of chatbots with semantic functionalities. Afterwards, I provided machine learning consulting from time to time.
+    description: Backend development for conversational AI (chatbots), followed by specialized Machine Learning consulting.
   - title: Research Intern
     company: University of Oxford
     company_url: ''
     company_logo: oxford
     location: Oxford, UK
     date_start: '2020-02-01'
     date_end: '2020-02-28'
-    description: Short-term research internship which lead to my BSc thesis about Temporal Knowledge Graph Reasoning.
-  - title: Teaching Assistant
-    company: Technical University of Vienna
-    company_url: 
-    company_logo: tu-vienna
-    location: Vienna, Austria
-    date_start: '2018-10-01'
-    date_end: '2020-09-30'
-    description: Teaching Assistant in *Algorithms and Data Structures* and *Fundamentals of Computer Engineering*
-  - title: Software Engineer Summer Intern
-    company: Iteratec GmbH
-    company_url: 
-    company_logo:
-    location: Vienna, Austria
-    date_start: '2019-06-01'
-    date_end: '2019-07-31'
-    description: Summer internship where I worked on a Flutter app and AWS integration
+    description: Research on Temporal Knowledge Graph Reasoning, which led to my BSc thesis.
+#  - title: Teaching Assistant
+#    company: Technical University of Vienna
+#    company_url: 
+#    company_logo: tu-vienna
+#    location: Vienna, Austria
+#    date_start: '2018-10-01'
+#    date_end: '2020-09-30'
+#    description: Teaching Assistant in *Algorithms and Data Structures* and *Fundamentals of Computer Engineering*
+#  - title: Software Engineer Summer Intern
+#    company: Iteratec GmbH
+#    company_url: 
+#    company_logo:
+#    location: Vienna, Austria
+#    date_start: '2019-06-01'
+#    date_end: '2019-07-31'
+#    description: Full-stack development (Flutter & AWS).
   - title: Industrial Automation Engineer
-    company: Jonas & Redmann GmbH
+    company: Keba AG / Jonas & Redmann GmbH
     company_url: ''
     company_logo: 
-    location: Berlin, Germany
-    date_start: '2016-01-01'
-    date_end: '2017-07-31'
-    description: Conceptualized and implemented software for industrial machines (solar industry) and put them to run on-site.
-  - title: Industrial Automation Engineer
-    company: Keba AG
-    company_url: 
-    company_logo:
-    location: Linz, Austria
+    location: Linz, Austria and Berlin, Germany
     date_start: '2014-05-01'
-    date_end: '2015-12-31'
-    description: Conceptualized and implemented software for industrial machines (plastics industry) and put them to run on-site.
+    date_end: '2017-07-31'
+    description: Software implementation and on-site commissioning for industrial machines (plastics & solar industry).
 design:
   columns: '2'
 ---
diff --git a/content/publication/2026-phd-thesis/index.md b/content/publication/2026-phd-thesis/index.md
@@ -0,0 +1,15 @@
+---
+title: "Accessible Time Series Forecasting with Deep Learning"
+date: 2025-12-01
+publishDate:  2025-12-01
+authors: ["**Andreas Auer**"]
+publication_types: ["2"]
+abstract: "Forecasting future values based on historical values is a critical task in many fields, from supply chain logistics and industrial production to energy management. Despite significant advances in deep learning based forecasting, practical deployment remains constrained by high operational overhead, including the need for initial training and periodic retraining. Additionally, many use cases lack sufficient task-specific data to train robust models. Furthermore, a forecast’s utility hinges on its reliability, which requires understanding of its inherent uncertainty — a feature often lacking in standard models. This thesis addresses this accessibility gap by developing methods aimed at enhancing the ease-of-use and reliability of forecasting without sacrificing performance. The contributions of this work are organized around two central pillars. The first pillar focuses on enhancing ease-of-use through zero-shot forecasting. By leveraging the paradigm of pre-trained models, we develop methods that can generate accurate forecasts across unknown datasets without requiring any task-specific training. This capability is achieved through in-context learning, where a model dynamically infers the patterns of a provided time series at inference time. We advance this paradigm by introducing COSMIC, a model that effectively incorporates external variables (covariates) in such a zero-shot setting — a crucial capability for real-world forecasting. This capability is instilled during pre-training via a novel augmentation strategy, and allows the model to dynamically infer covariate-target relationships. Further, we present TiRex, a pre-trained model based on xLSTM, which combines state-tracking with strong in-context learning, advancing the performance frontier of zero-shot forecasting. This is facilitated by a novel training paradigm that introduces Contiguous Patch Masking (CPM), alongside a suite of data augmentations, to force the model to rely on its internal state. The second pillar addresses reliability through uncertainty quantification. Recognizing that point forecasts are insufficient for decision-making, we adapt Conformal Prediction, a framework that provides guarantees with minimal assumptions, to the forecasting setting. Our proposed method, HopCPT, conceptualizes a time series as a sequence of recurring regimes and uses a Modern Hopfield Network to learn adaptive, context-aware uncertainty intervals. We extend this with HopCPT-G, which enables the transfer of uncertainty information across multiple related time series to improve predictions for rare events and data-scarce scenarios. In summary, this thesis advances the frontiers of zero-shot forecasting and uncertainty quantification under the unifying goal of accessible forecasting for a broad audience."
+featured: true
+publication: "Phd Thesis, JKU"
+links:
+  - icon_pack: fas
+    icon: download
+    name: Download
+    url: 'https://epub.jku.at/obvulihs/content/titleinfo/13227142 '
+---
