Adrian Calma
Senior Consultant
Experience
Senior Consultant
DB Systel GmbH
Spearheading the development of MLOps and crafting the MLOps strategies, guiding the market exploration efforts wrt. to MLOps, and managing requirements.
Key projects and contributions for DB Netz AG (08/2023 – 02/2024):
Product Owner MLOps
- Craft the architecture of the data flow, from data acquisition to model deployment
- Introduction of Scrum within the team
- Deployment of Kubernetes from scratch, on-premise
- Deployment of Kubeflow and setup of Prometheus for monitoring
- Development of a REST API with FastAPI for automatic triggering of pipelines
- Stakeholder management
- Writing and prioritization of user stories
- Tech stack: Kubernetes, FastAPI, Kubeflow, Jira, Confluence, Scrum
MLOps & ML Expert – Support in market exploration
- Evaluation of trials and test versions with a focus on MLOps and annotation
- Preparation of bidder discussions focusing on MLOps
- Support in various use cases with a focus on data management
- Tech stack: Kubeflow, CVAT, Jupyter Notebooks, Excel, Confluence, Jira, Kanban, PowerBI
Requirements Management – Data Factory
- Black box system requirements
- White box subsystem requirements for the ML platform
- Requirements analysis
- Requirements management
- Tech stack: Jira, Confluence, Polarion, PowerPoint, Excel
Consultant
AlphaPilots
Consulting for the MVP of the Airplane Price Evaluation Tool:
- Data selection, cleaning, and pre-processing
- Development of the machine learning model that estimates the price of an airplane
- Consulting regarding additional data sources
- Consulting regarding further ways of MVP improvement
- Consulting regarding feature usage tracking
- Tech stack: python, scikit-learn, numpy, pandas, google cloud functions, matplotlib
Product Owner
Signatrix GmbH
- Deployment of a completely new product within 2 weeks after the customer sketched the pain point
- Customer development, which lead to updated requirements for the self-checkout product
- Account management for a large retail customer
- Concept and writing of a grant proposal
- Pricing analysis
- Product ownership for the loss-prevention products, while developing a completely new AI-based product for the supervision of the baked goods shelves.
- Tech stack: Jira, Confluence
Venture Developer
German Deep Tech Institute
- Initial evaluation of over 90 potential business cases
- Market size & trends, competitor analysis, and overall evaluation of the potential for 6 selected business cases
- Generation of three new business cases
- Customer development, which lead to the first prototype
- Creation of the initial concept of the solution in figma, as clickable wireframe
- In depth market research, including market size, market trends, and competition analysis for 6 potential business cases.
- Tech stack: figma
Postdoc-UNIKAT-Fellowship
University of Kassel
The Postdoc UNIKAT Fellowship sponsored by the B. Braun Melsungen AG for outstanding applied research at the University of Kassel allowed me to explore the precision farming market and proof the concept with one potential client.
CEO – Tech and Business Development
Annotly UG (haftungsbeschränkt)
Business development and implementation of a SaaS/mobile app solution for early plant disease detection.
Key projects and contributions:
Mobile App for Precision Farming (01/2022 – 05/2022):
- Customer Development (over 50 customer interviews)
- Product Owner tasks: Product Backlog refinement, market analysis, customer interviews
- Developing a mobile app displaying a feed with alerts for farmers, where they can see which parcels/areas require their attention
- Creation of the initial concept of the app
- Management of the UI and its initial implementation
- Backend with nestjs and graphql
- Tech stack: javascript, react native, graphql, nestjs, expo, postgreSQL, figma, python, graphene, typeorm, docker
Proof of Concept for Matching RADAR Satellite Data to a Vegetation Index (12/2021 – 03/2022):
- Product Owner tasks: Product Backlog refinement
- Matching the geospacial data to the corresponding parcels
- Implementation of classification and regression algorithms for estimating the Normalized Difference Vegetation Index (NDVI)
- Supervising a student assistent that was helping with the task
- Tech stack: scikit-learn, numpy, pandas, geopandas, S3 bucket, PostgreSQL with PostGIS, python, SNAP
Internal REST API (12/2021):
- Internal REST API for communicating with the PostgreSQL database, including PostGIS functions
- Dockerized deployment as a microservice
- Tech stack: numpy, PostgreSQL with PostGIS, python, sqlalchemy, fastapi
Outlier Detection for Precision Farming (09/2021 – 11/2021):
- Determine the regions where plants are under stress using satellite data, the Normalized Difference Vegetation Index (NDVI) and geospacial information
- Implementation of an outlier detection algorithm
- Tech stack: scikit-learn, numpy, pandas, geopandas, PostgreSQL with PostGIS, python, SNAP, folium
Member of the Steering Committee
International Workshop and Tutorial on Interactive Adaptive Learning
Member of the Steering Committee of the International Workshop and Tutorial on Interactive Adaptive Learning, co-located with the European Conference on Machine Learning and Principles and Practice of Knowledge Discovery.
Proof of Concept for faulty package label detection
Annotly UG (haftungsbeschränkt)
- Customer Development
- Vision and User Stories for the prototype
- Object detection and binary classification of package labels: good or faulty
- Prototype implemented as a jupyter notebook, that was able to detect and classify them in a video stream
- Tech stack: scikit-learn, tensorflow object detection API, numpy, python
Senior Machine Learning Engineer – Human-in-the-Loop
Signatrix GmbH
Implement active learning in combination with deep learning; management of the annotation team.
Key projects and contributions:
Supervision and Optimization of the Annotation Process (06/2020 – 03/2021):
- Communication, consideration, and evaluation of the stakeholder’s requirements
- Evaluation and optimization of the internal annotation process
- Management of the annotation team
- Tech stack: python, pandas, CVAT
Exploratory Data Analysis and Data Cleaning (06/2020 – 02/2021):
- Exploratory data analysis of the existing dataset and data cleaning where it was necessary
- Evaluation of the annotation quality
- Dataset generation for deep learning tasks
- Tech stack: python, pandas, pytest, poetry, multiprocessing, shell scripts
Proof of Concept of a SaaS for organizing events
evesync
- Building a SaaS solution for potential clients that would like to manage the organization of an online event, from sponsors, to budgets, to speakers
- Customer development, business development
- Tech stack: vuejs, firebase
Podcast Host
Innovation Friday
A podcast dedicated to early stage entrepreneurs, those that are still fighting in the trenches. It is about their struggles, the most difficult, and most fulfilling moments as entrepreneurs.
Consulting for Human-in-the-Loop
Berlin based Startup
- Improving the Human-in-the-loop approach, by optimizing the data selection strategy and what/how the annotators are labeling the data
- Type of data: Video footage
- Tech stack: python, scikit-learn, pandas
Invited Speaker
Stanford Machine Learning Meetup
Introduction to Active Learning, co-located with JJLake’s 6am series. 45 minutes of explaining and motivating the use of active machine learning (human-in-the-loop) with 30 minutes hands-on.
Python Package for Simulating Annotators in Active Learning
University of Kassel
- Development of a python package that can simulate error-prone annotators
- Student supervision and task management
- Published under MIT License
- Tech stack: python, scikit-learn, pandas, numpy, matplotlib
Co-Founder – CTO
HI darwin Inc.
Implementing the decision support model for evaluating early stage investment opportunities.
Co-Founder – CTO
vencortex UG (haftungsbeschränkt)
The areas of responsibilities were a combination of CTO/CEO, focused more on the technical part.
Key projects and contributions:
Developing the MVP of the SaaS Platform for Guiding Startups and Evaluating Investment Opportunities (09/2018 – 03/2020):
- Database design, implementation and deployment of a web application that guides early stage startups from idea to product market fit (and seed investment)
- Implementation of the evaluation strategy
- Evaluating the algorithms with simple explainable AI approaches, e.g., counterfactual explanations
- Management of development team of up to 6 developers
- Tech stack: python, jupyter-lab, pandas, gerrit, shell scripts, sqlalchemy, angular, django, PostgreSQL
Knowledge Fusion of Multiple Expert Statements (11/2018 – 12/2018):
- Concept and implementation of an approach to fuse different, sometimes contradictory, expert statements.
- Tech stack: python, pandas, poetry, shell scripts, sqlalchemy
Lung Cancer Detection with Deep Learning
viscortex
- Proof of concept for the detection of malign nodules in CT scans by using a U-Net architecture
- Finalist of the Hessen Ideen Stipendium 2018
- Tech stack: python, keras, scikit-learn, pandas, matplotlib
Co-organizer
International Tutorial on Interactive Adaptive Learning
Co-organizer of the International Tutorial on Interactive Adaptive Learning, co-located with the IEEE World Congress on Computational Intelligence.
Ordinal Classification of Pre-Seed Startups
vencortex UG (haftungsbeschränkt)
- Concept and implementation of an ordinal classifier trained on data from pre-seed startups
- Tech stack: python, pandas, poetry, shell scripts, sqlalchemy
Co-organizer
International Workshop and Tutorial on Interactive Adaptive Learning
Co-organizer of the 1st, 2nd, 3rd, and 4th International Workshop and Tutorial on Interactive Adaptive Learning, co-located with the European Conference on Machine Learning and Principles and Practice of Knowledge Discovery.
PhD Candidate – Active Machine Learning
University of Kassel
Development of active machine learning algorithms that take the uncertainty of the annotators into consideration. Writing grant proposals and publishing 24 peer-reviewed papers.
Key projects and contributions:
Python Package for Simulating Annotators in Active Learning (01/2019 – 07/2019):
- Development of a python package that can simulate error-prone annotators
- Student supervision and task management
- Published under MIT License
- Tech stack: python, scikit-learn, pandas, numpy, matplotlib
Development of an Annotation Strategy with Self-assessed Confidence (01/2018 – 07/2018):
- Development of an annotation strategy that asks the annotators to provide self-estimated confidences and derives gradual labels
- Student supervision and task management
- Implementing the whole machine learning workflow: e.g., pre-processing, feature selection, design of the experiments, evaluation, annotation framework, etc.
- Tech stack: java
Collaborative Active Learning Web-App (10/2017 – 03/2019):
- Concept and development of a crowd sourcing web-app to label traffic signs within a predefined time and track the time and quality
- Student supervision and task management
- Deployment of the web-app and using it to label 10 thousand traffic signs
- Tech stack: python, django, angular, postgreSQL, docker
Development of a Selection Strategy for Active Learning
University of Kassel
- Development of a selection strategy for an active learning approach for classification tasks that make use of generative and discriminative models
- Implementing the whole machine learning workflow: e.g., pre-processing, feature selection, design of the experiments, evaluation, etc.
- Tech stack: java
Summary
Product Management and Ownership. Led cross-functional teams in the development and deployment of MLOps solutions, driving stakeholder alignment and prioritizing user stories for optimal product delivery.
Team Lead. CTO and CEO startup experience with teams of up to 9 persons (tech, sales, and marketing).
Mentoring and Consultancy. Technical guidance for the founders of iurCrowd, a Legal Tech Startup based in Berlin, on identifying resources and exploring development options for their MVP; spearheading the development of MLOps strategies and providing market exploration support for large German corporations.
Human-in-the-Loop (Active Learning). Longstanding experience with Human-in-the-Loop (Active Learning) approaches, including annotator management
Machine Learning and Pattern Recognition. Comprehensive knowledge with projects in object and outlier detection, and classification
Software Engineering (Python). Profound knowledge in the conception of software frameworks with realized applications in the areas of fintech and precision farming
Development and Operations (DevOps). Multiple years of running python microservices in a cloud environment (deployment, containerization, etc.)
Languages
Education
University of Kassel
PhD in Computer Science · Computer Science · Kassel, Germany
University of Kassel
Master of Science in Computer Science · Computer Science · Kassel, Germany
University of Kassel
Bachelor of Science in Computer Science · Computer Science · Kassel, Germany
Certifications & licenses
Certified SAFe® 6 Agilist
Professional Agile Leadership™ - Evidence-Based Management™ (PAL-EBM)
Professional Scrum Product Owner™ II (PSPO II)
Professional Scrum Product Owner™ I (PSPO I)
Similar Freelancers
Discover other experts with similar qualifications and experience