Benjamin Matschke-AI/ML/CV Engineer, System Architect, Founder, Mathematician
Check rate
Experience
Founder, system architect, and main developer
Institute for Artificial Study (IAS)
- Expert-supervised AI systems for scientific reasoning, model evaluation, and research workflows.
- Built the IAS Problem Solver, an orchestrated system for difficult mathematical reasoning; it achieved 84% in one submitted answer set on the Leipzig mathematics benchmark.
- Built a resumable state-machine pipeline for research-grade mathematics benchmark generation: source selection, LLM-agent-based phenomenon discovery, task synthesis, gold-answer and certificate generation and validation, probing, repair, human feedback, and quality gates, targeting tasks that are difficult, natural, verifiable, and cost-effective.
- Current work extends this into budget-aware AI research workflows for real scientific problems with expert review.
Tech stack: Python, OpenAI/OpenRouter-compatible APIs, embeddings, RAG, SQLite.
Founder, product architect, and main developer
better.town
- Pre-launch product R&D of AI-assisted tools for local communities, beginning with a structured mediation assistant for neighborhood conflicts.
- Designed and began implementing a structured workflow that helps users clarify the dispute, distinguish facts, positions, and interests, and explore possible agreements.
- Neutrality, privacy, safety, human escalation, and legal/compliance validation are treated as first-class product requirements.
Main developer
Project Manager Orchestrator (agentic devtools)
Python CLI orchestrating Codex via MCP with policies and quality gates.
- Built an orchestration layer that plans, executes, and reviews software tasks by driving a local coding worker via MCP.
- Implemented session reuse, streaming proposals, and controlled patch extraction/apply workflows.
- Tech stack: Python, MCP, Codex CLI.
Main developer
LLM billing gateway prototype (side project)
Gateway proxying OpenAI-style APIs with pricing/budgets and observability.
- Implemented an MVP API gateway that proxies upstream LLM requests and attaches cost/receipt metadata and budget enforcement hooks.
- Designed pricing configuration and monitoring (Prometheus metrics and tokenizer drift instrumentation) to support operating a paid API layer safely.
- Tech stack: Rust, Prometheus.
Main developer
Agentic Video Generator (side project)
Stage-aware pipeline for generating narrated videos.
- Designed a stage-aware workflow (prompt → plan/outline → draft → TTS → visuals → timeline → MP4) with resumable execution and explicit artifact boundaries.
- Implemented an event-sourced project model with JSON-patch updates to support arbitrary user edits, diffs, and resumable workflows.
- Built job execution with progress streaming (SSE/WS), retry/backoff, and pluggable seams for storage and visual providers.
- Tech stack: FastAPI backend, Next.js UI, FFmpeg rendering.
Primary technical owner, Principal ML/CV Engineer
WWE Media GmbH
Video analysis from football broadcasting:
- computer vision: keypoint and camera detection (from a single camera); player and ball detection, instance segmentation, and tracking; optimal global tracks; human pose estimation and classification; stitching images from same and different cameras; camera synchronisation (recognizing same scene from different camera angle); text recognition and player reidentification; predict 3D ball trajectory from single camera; everything via real-time/online algorithms
- machine learning: role classification, event detection (rare events such as penalty kicks, red cards); compute team statistics from video; audio processing (whistle detection); game status detection from player detections; ball position estimation from player detections
- methods: various deep neural networks, Kalman filters, clustering feature vectors, rare event detection (decision trees, hidden Markov models, auto-encoders), constraint programming, photogrammetry (SIFT, Hough transform), bundle adjustment, fitting physics models; optical flow
- tech stack (condensed): Python (PyTorch, OpenCV), ffmpeg; FastAPI microservices; Celery/Redis and NATS; Ray + Optuna; Docker, GitLab
My role: primary technical owner; collaborated with the CEO and coordinated with a larger ML/CV team early on, plus annotators, occasional frontend support, and academic collaborators (University of Hamburg).
Event Manager
Boston University
Organized Big Data in Pure Mathematics 2022, an international online workshop at Boston University bringing together leading developers and researchers from mathematical databases and knowledge systems such as LMFDB, OEIS, Lean/mathlib, FindStat, polyDB, and Fungrim. The workshop addressed search and accessibility, data quality, long-term preservation, interoperability, and collaborative infrastructure for large-scale mathematical data.
Founder and main developer
NumberDB
Founder and main developer of NumberDB, a scientific platform for identifying and exploring mathematical constants and polynomials. Designed and built the core database, search and identification algorithms, computational backend, public API, Python package, AI integration skill, web application, and deployment infrastructure using Python/Django, PostgreSQL, SageMath, PARI/GP, Arb, nginx, Gunicorn, Pyro5, Docker, and GitHub Actions.
Main developer
Private project
Desktop application (Python) to help create photo-realistic paintings.
Research Assistant Professor
Simons Collaboration on Arithmetic Geometry, Number Theory, and Computation (Boston University, funded by Simons Foundation)
Research Assistant Professor in computational mathematics and number theory; developed algorithms and software for large-scale mathematical computation, contributed to LMFDB, founded NumberDB, and worked on mathematical data mining, database infrastructure, search, deployment, and scientific cloud computing.
Collaborator
LMFDB
A database in number theory. Written in Python (Flask), using MongoDB and PostgreSQL. I was one of around 50 developers. We did regular agile coding sprints (scrum-like).
Assistant Professor in Mathematics
Koc University
Teaching: Probability Theory, Statistics, Discrete Mathematics, Complex Analysis, Algebra I, Calculus, Topics in Topological Combinatorics. Co- supervised 2 Master's students.
Main developer
Private project
Python classes for discrete Riemann surfaces.
Principal Investigator
University of Bordeaux
Principal Investigator of project “Discrete Geometry Methods in Algorithmic Number Theory” (funded by Excellence Initiative of University of Bordeaux).
Main developer
Private project
Some audio tools (Java): Measuring pitches of voice, virtual instrument for harmonies, etc.
Main developer
Private project
Python classes for circle packings.
Main developer
Private project
Research article “Computing integral points on $X_{ns}^{+}(p)$”, with Aurélien Bajolet and Yuri Bilu, Algebra Number Theory 15 (3), 2021, 569-608, source code (Python/SageMath) and data available.
Main developer
Private project
Mathematical Java-applets.
Research Assistant
TU Berlin
Research Assistant at Research group “Polyhedral Surfaces” (funded by DFG).
Postdoc
Many institutes
At:
- Institute for Advanced Study (Princeton)
- Institut des Hautes Etudes Scientifiques (Paris)
- ETH Zürich
- Isaac Newton Institute (Cambridge)
- Max Planck Institute for Mathematics (Bonn)
- Max Planck Institute for Mathematics in the Sciences (Leipzig)
- University of Bordeaux
- Technion (Haifa) Research in discrete geometry, linear optimization, algebraic topology, combinatorics, number theory; scientific cloud computing
Main developer
Private project
A decent ray-tracer (C++), including backwards-ray-tracing.
Main developer
Private project
A programmable virtual robot (Delphi), which is a Turing complete programming language that is aimed to be simple and accessible to any high school student.
Main developer
Private project
Simulations of basic mechanical systems (Delphi).
Main developer
Private project
A Delphi package for big integers and some cryptography (RSA in Delphi).
Main developer
Private project
A web crawler (PHP+JS).
Main developer
Private project
Function parser and plotter (Delphi).
Main developer
Private projects
Various desktop applications and games (Delphi, C/C++, Pascal, QBasic), including board games with AI (chess, Abalone, 4-Connect), a 3-dimensional Snake, 2D and 3D jump'n'runs, multi-player network games.
Industry experience
See where this freelancer has spent most of their professional time.
Experienced in Information Technology, Education, Media and Entertainment, Sport, Arts and Crafts, and Manufacturing.
Business area experience
See which departments and functions this freelancer has contributed to most.
Experienced in Information Technology, Product Development, Research and Development, Project Management, and Finance.
Summary
Principal ML Engineer & Technical Owner with a PhD in Mathematics, specializing in production-grade AI systems with a strong focus on real-time machine learning and computer vision. I build robust systems end-to-end — from problem framing and evaluation to model development, deployment, and stable operation.
In recent years, I owned the technical delivery of a real-time football video analytics platform (single-camera and broadcast), covering pitch geometry, player and ball tracking, and automated event and statistics extraction. The system was designed for reliability, measurable performance, and real-world constraints rather than demos.
I enjoy working on technically hard problems that require both deep ML expertise and system-level thinking. Fully remote. Open to permanent roles or long-term B2B contracting.
Skills
- Software Architecture
- Technical Leadership
- Ai/Ml Technical Ownership
- End-To-End Ml System Design
- Production Machine Learning
- Computer Vision
- Video Analytics
- Real-Time Video Analysis
- Multimodal Ai Systems
- Deep Learning
- Generative Ai
- Llm Systems
- Agentic Ai Systems
- Ai Agents
- Scientific Reasoning Systems
- Model Evaluation
- Ai Output Evaluation
- Human-In-The-Loop Ai
- Expert-Supervised Ai Workflows
- Prompt Engineering
- Prompt Authoring
- Resumable Agentic Workflows
- Budget-Aware Ai Orchestration
- Automated Validation And Repair
- Quality Gates For Ai Systems
- Llm Api Orchestration
- Llm Api Gateways
- Llm Cost Tracking And Budget Enforcement
- Model Observability
- Computer Vision System Architecture
- Object Detection
- Small-Object Detection
- Player Detection
- Ball Detection
- Instance Segmentation
- Semantic Segmentation
- Object Tracking
- Multi-Object Tracking
- Small Fast Object Tracking
- Global Track Optimization
- Player Tracking
- Ball Tracking
- Person Re-Identification
- Player Re-Identification
- Human Pose Estimation
- Human Pose Classification
- Camera Detection
- Camera Pose Estimation
- Camera Calibration
- Photogrammetry
- Bundle Adjustment
- Image Stitching
- Multi-Camera Image Stitching
- Camera Synchronization
- Cross-Camera Scene Matching
- 3d Geometry
- 3d Reconstruction
- 3d Ball Trajectory Estimation
- Physics-Based Trajectory Fitting
- Optical Flow
- Keypoint Detection
- Feature Matching
- Sift
- Hough Transform
- Ocr
- Text Recognition
- Audio Processing
- Whistle Detection
- Event Detection
- Rare-Event Detection
- Sports Analytics
- Soccer Analytics
- Team-Statistics Computation
- Game-State Detection
- Role Classification
- Time-Series Modeling
- Kalman Filtering
- Hidden Markov Models
- Decision Trees
- Autoencoders
- Clustering
- Feature-Vector Clustering
- Statistical Modeling
- Optimization
- Linear Optimization
- Polynomial Optimization
- Semidefinite Programming
- Mixed-Integer Programming
- Constraint Programming
- Computational Geometry
- Computational Complexity
- Algorithm Design
- Efficient Algorithms
- Numerical Algorithms
- Scientific Computing
- Computational Mathematics
- Computational Number Theory
- Discrete Geometry
- Algebraic Topology
- Combinatorics
- Number Theory
- Arithmetic Geometry
- Diophantine Equations
- Elliptic Curves
- S-Unit Equations
- Lattice-Point Methods
- Mathematical Research
- Mathematical Problem Solving
- Mathematical Data Mining
- Mathematical Databases
- Scientific Database Design
- Big Data
- Data Mining
- Pattern Discovery In Mathematical Data
- Conjecture Discovery From Data
- Scientific Cloud Computing
- Python
- Pytorch
- Opencv
- Fastapi
- Flask
- Django
- Next.Js
- Rust
- Java
- C++
- C
- Pascal
- Delphi
- Php
- Javascript
- Sagemath
- Pari/Gp
- Arb
- Ffmpeg
- Onnx
- Tensorrt
- Ray
- Optuna
- Celery
- Redis
- Nats
- Mcp
- Codex Cli
- Prometheus
- Postgresql
- Mongodb
- Minio
- Docker
- Nginx
- Gunicorn
- Pyro5
- Git
- Github
- Gitlab
- Github Actions
- Ci/Cd
- Microservices
- Rest Apis
- Web Apis
- Backend Development
- Full-Stack Web-System Design
- Database Architecture
- Data Pipelines
- Ml Pipelines
- Streaming Systems
- Server-Sent Events (Sse)
- Websockets
- Event-Sourced Architectures
- Json Patch
- Retry And Backoff Mechanisms
- Workflow Orchestration
- Distributed Job Execution
- Cloud Deployment
- Model Deployment
- Production Monitoring
- Health Checks
- Latency Optimization
- Compute-Cost Optimization
- Resource-Efficient Ml
- Scalable Ai Systems
- Robust Production Pipelines
- Api Design
- Research-To-Production Translation
- Prototype-To-Production Delivery
- Technical Strategy
- Product-Oriented Ai Development
- Cross-Functional Collaboration
- Team Coordination
- Ml Engineer Management
- Annotator Coordination
- Academic Collaboration
- Research Mentoring
- Student Mentoring
- University Teaching
- Conference Organization
- Scientific Communication
- Research Presentations
- Research Paper Writing
- Grant-Funded Research Leadership
- German (Native)
- English (Fluent)
- French (B2)
Languages
Education
FU Berlin
Ph.D. in Mathematics · Mathematics · Germany · summa cum laude
TU Berlin
Diplom in Mathematics · Mathematics · Germany · 1.0
Max-Steenbeck-Gymnasium
Abitur · none
Statistics
Experience
Global experience
Expertise
Qualifications
Profile
Frequently asked questions
Have questions? Find more information here.
Benjamin is based in Cottbus, Germany.
Benjamin speaks the following languages: German (Native), English (Advanced), French (Intermediate), Turkish (Elementary).
Benjamin has at least 29 years of experience. During this time, Benjamin has worked in at least 12 different roles and for 16 different companies. The average length of individual experience is 1 year and 1 month. Note that Benjamin may not have shared all experience and actually has more experience.
Based on recent experience, Benjamin would be well-suited for roles such as: Founder, system architect, and main developer, Founder, product architect, and main developer, Main developer.
Benjamin's most recent position is Founder, system architect, and main developer at Institute for Artificial Study (IAS).
In recent years, Benjamin has worked for Institute for Artificial Study (IAS), better.town, Project Manager Orchestrator (agentic devtools), LLM billing gateway prototype (side project), and Agentic Video Generator (side project).
Benjamin is most experienced in industries like Information Technology, Education, and Media and Entertainment. Benjamin also has some experience in Sport, Arts and Crafts, and Manufacturing.
Benjamin is most experienced in business areas like Information Technology, Product Development, and Research and Development. Benjamin also has some experience in Project Management and Finance.
Benjamin has recently worked in industries like Information Technology, Media and Entertainment, and Sport.
Benjamin has recently worked in business areas like Information Technology, Research and Development, and Product Development.
Benjamin holds a Doctorate in Mathematics from FU Berlin and a Bachelor in Mathematics from TU Berlin.
Benjamin is immediately available full-time for suitable projects.
Daily rate distribution
The rates shown represent the typical market range for freelancers in this position based on recent contracts on our platform.
Average rates for similar positions
Rates are based on recent contracts and do not include FRATCH margin.
The average daily rate is the mean of all daily rates from recent contracts of comparable freelancers on our platform.
The median daily rate is the middle value of all daily rates — half of comparable freelancers charge less, half charge more. Unlike the average, it is barely affected by outliers.
Calculated based on our freelancers’ daily rates as of 30 Aug 2026. Actual rates may vary depending on seniority level, experience, skill specialization, project complexity, and engagement length.
Similar freelancers
Discover other experts with similar qualifications and experience
Tobias Jaeuthe
Jürgen Fey
Louis Guitton
Alexander Lechner
Ulf Schiebener
Mathias Wilhelm
Thomas Hoefkens
Yannick Schuchmann
Arne Hendricks
Danny-Michael Busch
Experts recently working on similar projects
Freelancers with hands-on experience in comparable project as a Founder, system architect, and main developer
Laurin Hagemann
Matthias Isler
Manuel Pasieka
Marcel Meyer
Rutger Boels
Felix Schaller
Kai Schmudlach
Anton Klonov
Fouad Omri
Dennis Dickmann
Nearby freelancers
Professionals working in or nearby Cottbus, Germany
Christel Kristin Riedel
Frank Noack
Andreas Rühl
Silvia Bürmann
Chris Wolf
Dmitry Pankov
Chintan Padaliya
Björn Stark
Hanna Fischer
André Siegmund
