My_CV.doc
Projects
Internet Explorer
DOOM
Media Player
Notepad
Calculator
Paint
LinkedIn
GitHub
Clippy Chat
Status: (loading...)
Ask Clippy a question.
Johannes
WordPad
My Projects
DOOM
Internet Explorer
Media Player
Notepad
Calculator
Paint
My Documents
My Pictures
My Music
My Computer
Control Panel
Turn Off Computer
10:45 PM
Volume
My_CV.doc - WordPad
Johannes Janousek
johannes.janousek@gmail.com | +43 6645457202 | Vienna
linkedin.com/in/j-janousek | github.com/jjanousek | johannesjanousek.xyz
Experience
Penxion
Jan 2024 - Dec 2025
Technical Founder
Berlin
- Built a platform helping seniors and their caretakers with care related tasks in Germany.
- Developed AI-driven web crawlers to compile Germany's largest dataset on senior homes, utilizing multi-modal LLMs to analyze images and extract features.
- Engineered data pipelines using Airflow on AWS to process, clean (using Pandas, NumPy), and store crawled and application data across S3, RDS (MySQL, PostgreSQL), and MongoDB.
- Created backend tools (using Python, JS, Flask, FastAPI) enabling AI components to generate custom care home reports.
- Implemented functionality to assist seniors in filling out crucial forms for Pflegegrad assistance.
- Integrated frontend and backend systems via webhooks connected to CRM and EC2 instances for triggering automated processes.
Massar Capital (HFM US
Performance Award 2023)
Aug 2022 - Jan 2024
Data Scientist
Vienna
- First hire in newly opened Vienna office; worked directly with CIO across research, data infrastructure, and trading analytics.
- Developed a predictive model optimizing the roll dates for >100 illiquid commodity futures, saving $10 million annually in slippage costs.
- Researched and developed models to predict surprises in economic releases.
- Engineered systems for acquiring, processing, and validating large datasets from exchanges, vendors, Bloomberg Terminal and SQL databases (MySQL, MSSQL) to feed models and analytics.
- Acquired and prepared alternative datasets, including satellite imagery for oil inventory analysis, to support trading strategies.
- Built and maintained web crawlers and monitoring systems with alerts for time-sensitive trading strategies.
- Designed and implemented an interactive analytics tool (Dash, Plotly) enabling the CIO to analyze seasonality trends and translate research into trades.
WeMine
Dec 2019 - Jul 2022
Machine Learning Engineer
Hong Kong
- Developed and deployed a multi-task deep learning recommendation system (TensorFlow, PyTorch) that incorporated user context to drive engagement.
- Engineered scalable, low-latency prediction services on Azure using Docker and Kubernetes (AKS).
- Designed and automated end-to-end ML pipelines with Airflow, orchestrating data ingestion, preprocessing, training, and deployment.
- Systematized ML experimentation and model versioning using MLflow and CometML.
- Built custom Docker images to minimize image size and significantly lower cloud hosting costs.
ECO Capex
Jan 2016 - Sep 2017
Quantitative Analyst
London
- Created a credit risk analysis tool in Python and built predictive models to forecast trade success and revenue for a B2B digital marketplace.
- Designed and managed a currency basket based on volatility analysis, presenting key performance indicators and analytical outputs to the C-suite using Tableau.
Projects
Task Adaptive LoRAs
Instant Transformer Adaptation with DSPy | GitHub
- Re-implemented SakanaAI's Text-to-LoRA research paper with DSPy, generating task-specific adapters in one forward pass and optimizing prompts for the hyper network.
Education
King's College London
Sep 2018 - Sep 2019
MS Data Science and Statistics
Distinction
- Dissertation: Using convolutional neural networks to classify brain-computer-interface EEG data of epilepsy patients.
- Modules: Pattern Recognition, Deep Learning and Neural-Networks, Machine Learning, Artificial Intelligence, Databases, Data Warehousing, Statistics.
General Assembly London
Oct 2017 - Feb 2018
Data Science Course
Regent's University London
Feb 2013 - Jan 2016
BS Finance
Distinction
- Dissertation: Valuation Model for Young Tech Companies.
Other
- Nationality: Austria
- Languages: German, English
- Interests: Reading, Hiking, Diving, Skiing, Running
Projects
1 object
dspy-t2l
Task-Adaptive LoRAs with DSPy
Wraps Sakana AI's Text-to-LoRA hyper-network in a DSPy module to generate task-specific
LoRA adapters from plain-English task descriptions.
Stack: Python, DSPy, PyTorch, PEFT
Core module:
TaskToLoRA
Eval tasks: GSM8K and MedMCQA
Cannot find server - Microsoft Internet Explorer
File
Edit
View
Favorites
Tools
Help
Address
If the page doesn't load securely, click here to open in a new tab.
The page cannot be displayed
The page you are looking for is currently unavailable. The Web site might be experiencing technical difficulties, or you may need to adjust your browser settings.
DOOM
Windows Media Player
00:00
Playlist
Untitled - Notepad
File
Edit
Format
View
Help
Calculator
untitled - Paint
AI
Enhancing...
AI ✨
Size
For Help, click Help Topics on the Help Menu.
My Music
Available Tracks:
Kokoroko
Reggae
Harvest Sky
Obvious
My Pictures
Available Pictures:
Bliss.jpeg