Software development and training for technical teams
I offer in-house training for development teams, built around their own code, tools, and workflows.
I build custom software for difficult technical problems, such as mathematical optimization and numerical computing, when standard solutions don't fit.
Write to me at baharev@bitalon.com in English or German for a first meeting, in person or online, with no obligation.
- PhD in chemical engineering; modeling and simulation
- 12 years of research in applied mathematics
- Since 2019 self-employed software developer; trainer since 2022
- 59,029 reputation points on Stack Overflow
Training for your development team
My training is for developers, engineers, and scientists who write code. Public courses follow a preset curriculum; I build in-house training around your team's code, tools, level, and goals, and deliver it in English or German, on-site (in Austria or abroad) or online. It is not a general introduction to ChatGPT: every course works on what your team actually needs.
The process is simple: a first meeting, where you tell me the topics you need or we work them out together from your code and tools; a proposal; the training; and a follow-up call about two weeks later. If you wish, I can also support your team with coaching hours while it applies what it learned to its own project.
Training in AI-assisted software development can also contribute to your team's AI literacy under the EU AI Act.
You benefit if
- your developers want to use AI tools such as Claude Code or GitHub Copilot on your own code, with clear rules for review and testing
- public courses are too general for your codebase, your tools, or the mix of levels in your team
- you already know which topics your team needs, perhaps parts of several standard courses, and want exactly those, without the rest
- you want the training in English or German, on-site or online, on dates that suit your project
Training topics
- AI-assisted software development: Claude Code, GitHub Copilot, and similar tools on your own codebase, with code review, testing, and validation
- AI in your own applications: language model APIs (Claude, GPT, Gemini) with tool use and agents, and MCP servers that connect your tools and devices to AI models
- Python for engineers and scientists: numerical methods, mathematical optimization, data analysis, automation
- Full-stack web development: the MERN stack of MongoDB, Express, React, and Node.js
- Software engineering: requirements, design, Git workflows, testing, and code review, rethought now that AI coding agents write a growing share of the code
- Modern C++: current language features and the standard library, for performance-critical and numerical code
- React and React Native: component-based user interfaces for the web, and mobile apps for Android and iOS on the same basis
- SQL and database design: from first queries to the design of relational databases
What participants say
Extremely likable, empathetic, and fair. Ali made a real effort to ensure we truly understood the material instead of forcing us through it. He actively involved us in the lessons.
Broad topics explained concisely and explored in depth with practical examples. It was great, thanks.
Custom software development
I build software for problems that require more than a standard application: algorithms, numerical methods, mathematical optimization, statistics, time series analysis, machine learning, AI integration, data processing, and automation. Programming languages: C++, Python, JavaScript, Java, SQL.
I also work as a specialist inside existing projects, alongside your team, in Python or C++.
For new software, we start with a small, fixed-price piece of work, so both sides can assess the collaboration before committing to more.
You benefit if
- a key calculation lives in a spreadsheet that only one person understands
- planning or scheduling is still done by hand, and every change costs hours
- a simulation, analysis, or mathematical optimization model runs too slowly, or its results are hard to trust
- no standard product fits the problem, and your own team lacks the time or the specialist background
Examples from projects and research
- Large-scale mixed-integer linear programming (MILP): custom branch-and-bound solver for a complex industrial scheduling problem
- AI visual inspection over MCP: an AI model inspects parts in camera images (my own open-source project on GitHub)
- Quantitative finance: analysis and comparison of stock and bond portfolio performance, scenario simulation, and time series analysis
- Machine learning for industrial process analysis: which of several hundred process parameters determine a product quality characteristic with a narrow tolerance
- Statistical analysis for a dissertation on aphasia: 54 participants in four groups; all figures and tables of the thesis's 70-page results chapter, generated automatically from the data
- Geological data analysis: a C++ and Qt desktop program with regression and cluster analysis, developed with a geologist and published in Computers & Geosciences
- Wearable motion sensor for a medical application: microcontroller firmware recording at about 200 Hz, sensor fusion using nonlinear programming instead of a Kalman filter, and 3D visualization in C++ and OpenGL
- Graph algorithms and decomposition methods: exact method for the minimum feedback arc set problem; decomposition of large systems of equations in process simulation into a sequence of small subsystems
- Full-stack web applications: my own products for speech therapy after a stroke (paying customers on subscription since 2021) and for memory games; React with a serverless back end on AWS Lambda and DynamoDB
For software companies
I take on work that doesn't fit your usual projects or is too specialized to staff in-house: hard algorithms, mathematical optimization, numerical and performance-critical code in C++ and Python, and technical spikes. I can also train your developers on your own code. I work behind the scenes and do not approach your clients.
Experience and qualifications
Background
- PhD in chemical engineering: numerical computations in thermodynamics, fluid phase equilibria, and distillation columns
- Applied mathematics: 12 years of research in numerical methods and mathematical optimization (2006–2018), 7 of them at the University of Vienna
- Self-employed since 2019 (Bitalon IT Services e.U.): software development in mathematical optimization, algorithms, numerical methods, and data processing
- Software trainer since 2022: WIFI, other training providers, and in-house; from beginners to senior developers
- Full-stack web development since 2018: React, Node.js, and serverless AWS
- Programming: C++, Python, JavaScript, Java, SQL
- Sun Certified Java Programmer (310‑065)
- Oracle Database 11g SQL Certification (1Z0‑051)
Public profiles
- LinkedIn: professional experience
- XING: professional experience, in German
- GitHub: open-source projects
- Stack Overflow: 59,029 reputation points, 5 million people reached
- Google Scholar: scientific publications
Frequently asked questions
Can we use AI tools on confidential code?
Usually yes, with the right setup. Before the training, we go through your company's rules; I then show which tool settings and subscription plans comply with them, for example business plans under which the provider does not train its models on your code. If your rules exclude cloud services, we discuss what is possible with models on your own infrastructure.
Do you use AI tools in your own development work?
Yes, every day, mainly Claude Code, including on production software for paying clients. I set the architecture, the rules, and the tests; the AI writes code within them, and every change is reviewed and tested. My training is built on this way of working.
How do you tailor the training to our code?
From the first meeting and a look at your code and tools, I learn what your team builds and where it gets stuck. The exercises then work on your own codebase, or on a realistic part of it, so the team practices on the problems it actually has.
What if a standard product would do?
Then I tell you so, and you save the money. Custom software pays off only when no product fits the problem.
What rights do we get to the code?
You receive the full source code under the license we agree on in the contract, typically the MIT license.
How do you set the price?
Training: one flat price per team per day, which I quote after the first meeting. Custom software: a new project starts with a small piece of work at a fixed price; after that, we agree on the next steps and their price together.
Get in touch
Write to me at baharev@bitalon.com in English or German for a first meeting, in person or online, with no obligation.