Open to Summer 2027 internships · Frankfurt am Main, Germany

Software engineering for financial systems.

I’m Max, a Computer Science Master’s student at DePaul University, focused on artificial intelligence, software engineering, and computational finance. During my time at DekaBank, I developed an algorithm that maps thousands of financial news articles to issuers and client portfolios, now moving toward production deployment.

I’m currently looking for a Summer 2027 internship in AI engineering, software engineering, or fintech.

About

I’m a computer scientist interested in engineering intelligent systems, developing algorithms, and solving complex problems through software.

I’m pursuing a Master’s in Computer Science at DePaul University, coming from a finance background, and I build software and AI systems for financial applications. My work spans AI engineering, computational finance, and systems: a fully local RAG pipeline with an evaluation harness over SEC 10-K filings, and an economic forecasting dashboard whose models are chosen by backtest.

At DekaBank’s Structuring & Derivatives Trading unit, I built a Python algorithm that matches thousands of incoming news articles to issuers and client portfolios. It’s now being moved into production. I also built a rule-based Java classification engine, evaluated an LLM-based term-sheet comparison tool, prototyped an interactive world map for the internal trading dashboard and presented it to senior engineers, and documented AI-assisted and agent-based development workflows.

I speak German, English, and Polish.

Selected projects

Things I’ve built.

01

Local RAG & Evaluation Pipeline

A fully local RAG pipeline over FY2025–26 10-K filings of 5 companies (Microsoft, NVIDIA, Tesla, Lockheed Martin, BlackRock), using hybrid dense + BM25 retrieval with no external API calls. Oracle tests showed generation, not retrieval, was the bottleneck.

0.87
Answer score on 65 hand-verified questions, vs. 0.24 without retrieval (95% bootstrap CIs)
0.92
Retrieval hit@5

Python · MLX · Llama 3.1 8B · Chroma

View on GitHub ↗
Report summary: 0.87 answer score with RAG, 0.24 without retrieval, 0.92 retrieval hit@5 Report page on the oracle test: perfect retrieval doesn't raise the score Terminal output of the oracle test run, oracle vs. RAG score per question

02

World of Data

A live economic and geopolitical dashboard with forecasts computed in the browser and an AI analyst that answers by calling 27 live-data tools (World Bank, EIA, ACLED, market quotes), so it explains measured numbers instead of inventing them. Each model was chosen by a rolling-origin backtest across 217 economies (1990–2025) with a rule written down before the run: AR(2) won on GDP but lost on inflation and unemployment, which now ship a no-change forecast. IMF forecasts still beat AR(2) in 142 of 185 economies. Weekly US gas-storage forecasts beat the five-year seasonal norm by 5% over 563 weeks (Diebold–Mariano p = 0.004).

8% / 20%
Lower GDP error than “no change”, one and three years ahead
93–96%
Of 2010–2025 outcomes inside 95% intervals calibrated on pre-2010 forecasts

TypeScript · React · World Bank / EIA / ACLED APIs · LLM tool use

View on GitHub ↗
6 more screenshots

03

NEOTracker (side project, astronomy)

I’ve been into astronomy for a long time, so I built the app I always wanted when looking up at night. Point your phone at the sky and it tells you what you’re looking at: a planet, a star, a galaxy or an asteroid. It uses your GPS location and the phone’s orientation, and I compute the planet positions on the phone with Meeus/VSOP87.

It also tracks near-Earth asteroids from public NASA and JPL data. For each one it shows a simple risk score I designed myself, and I explain in the app exactly how it’s calculated. I added a Science Lab with an orbit playground, an impact-energy calculator and classroom activities, because I think the physics is the most interesting part.

Coming soon to the App Store

Swift · SwiftUI · SceneKit

NEOTracker sky view with Jupiter selected, showing its distance and constellation NEOTracker sky view tracking near-Earth asteroid 2018 VA2 over Chicago NEOTracker orbit playground with sliders for an asteroid's orbit
5 more screenshots
NEOTracker launch screen with a star globe built from 25,791 real stars NEOTracker list of this week's close approaches with distance, velocity, diameter and risk score NEOTracker impact energy calculator showing the math step by step NEOTracker Explore Space menu with news, sky, constellations and 3D models NEOTracker space news feed with date filters

Experience

Where I’ve worked.

  1. Jun — Sep 2026Frankfurt

    Intern, Structuring & Derivatives Trading

    DekaBank

    • Developed a regex-based Python algorithm that processes thousands of dpa news articles and maps them to issuers and client portfolios, enabling scalable classification by issuer, category and country; currently being migrated into production
    • Developed a rule-based Java engine for the automated classification of counterparties under the EU Capital Requirements Regulation (CRR), with a test suite validating edge cases and rule priorities
    • Code-reviewed and performance-tested an LLM-based term-sheet comparison tool, assessing the quality and accuracy of its results; validated and documented test results in Excel (Python)
    • Developed an interactive world map prototype for the internal trading dashboard using Claude (TypeScript), combining internal financial data with external economic indicators; presented it to senior engineers as a demonstration of AI-assisted development in a regulated banking environment
    • Prepared internal wiki documentation on AI-assisted and agent-based coding workflows, using the world map as the case study
  2. Jun — Aug 2025Berlin

    Intern, Financial Reporting & Systems

    SWOT Controlling GmbH

    • Consolidated financials across five group companies and several external clients into standardized P&L reports
    • Built account mapping frameworks and automated Excel-based workflows to streamline financial data integration
    • Translated the firm’s controlling software from German to English, creating structured implementation files for the software engineers
    • Standardized reporting processes in a customer implementation project; supported system integration and client rollout alongside the CEO
    • Managed internal documentation in GitBook with version control through GitLab, and redesigned the user handbook
  3. May — Aug 2024Berlin

    Intern, Tax Accounting & Audit

    Schwanke & von Schwerin

    • Assisted senior accountants in preparing and reviewing financial statements for clients
    • Supported internal and external audits with audit preparation and document reviews, reducing filing errors by 15%; flagged 10% budget cost variances
    • Automated Excel checks that caught input errors from company software, improving data collection and reporting workflows
    • Researched compliance standards and best practices

Education & athletics

Where I’ve studied.

  1. Expected March 2028Chicago

    MS Computer Science

    DePaul University

    • GPA 3.8 / 4.0
    • Research interests: trading, machine learning for financial markets, AI engineering
    • Data Structures I–II, Computer Systems I–II, Discrete Mathematics, Python Programming, Distributed Systems, Applied Algorithms
    • DePaul Computer Science Society
  2. Graduated May 2025New Orleans

    BS Finance, Data Analytics Concentration

    University of New Orleans

    • Athletic and academic scholarship
    • Student-athlete on the NCAA Division I men’s golf team, Southland Conference
    • Privateer Blue Honors — Fall 2022, Spring 2023, Spring 2024
    • Investments, Financial Modeling, Quantitative Methods, Data Analytics, Accounting
    Golf roster: UNO Privateers ↗
  3. Aug 2020 — May 2022Wichita, Kansas

    BA International Business

    Wichita State University

    • Athletic and academic scholarship
    • Member of the Wichita State men’s golf team; NCAA All-Academic Team, 2021–22
    • Volunteered at The Lord’s Diner, serving meals in the Wichita community
    • International Business, Principles of Macroeconomics, Introductory Business Statistics, Business Software: Excel
    Golf roster: Wichita State ↗
  4. 2018 — 2020Germany

    German National Golf Team

    Competed for the German National Golf Team from 2018 to 2020. Native German speaker, fluent in English and Polish.

    In the press: Berliner Zeitung ↗

Skills

Skills & interests.

Programming

Python, Java, TypeScript, C, SQL (SQLite)
RAG (Chroma, BM25, MLX, Llama 3.1)
Tools: Git, Claude Code

Business

Excel (financial modeling)
PowerPoint
Bloomberg Terminal

Languages

German (native)
English (C2)
Polish (C2)

Beyond the desk

NCAA D1 Golf (5 years)
German National Golf Team (2018–2020)
Golf, geopolitics, cosmology, trading

Contact

Let’s talk.

Looking for a Summer 2027 internship.

mwojcie4@depaul.edu

Downloads

For friends.

Tools I built for friends. Free to download.

Mac app

CV Drafter

I built CV Drafter for friends and kept running into the same problem with AI-written CVs: they sound good, but they make things up. So I designed it around one rule: it can only use true facts about you.

You paste a job posting and get a CV and cover letter tailored to it. Every bullet and paragraph cites the facts it came from. Rule checks in code and a separate AI fact-checker verify each statement, and nothing gets rendered until the check passes. Names, dates and the document design are locked to your profile and a Word template, so the AI can only select and reword, never invent.

I also wanted to know which postings were worth applying to, so each one gets a fit verdict from fixed scoring logic, not from the AI. Every application is logged with its status and a follow-up reminder, and your edits become style rules for the next draft. Everything stays on your Mac, and it runs on your Claude Pro or Max plan with no API key.

9
Honesty safeguards, from cited facts to a render gate that blocks unverified text
0–100
Fit score per posting, from fixed logic you can inspect

Python · Streamlit · Claude Sonnet & Haiku · docxtpl · SQLite

Download for Mac View on GitHub ↗ macOS 12+ · Apple Silicon & Intel

Mac app

Gesture Control

I built Gesture Control because I wanted to see if I could move windows around my Mac without touching the trackpad, using only the camera that’s already built in. Hold up an open palm to start, pinch to click, pinch and hold to carry a window across all your screens, and open your hand to drop it. A tight fist minimizes the window under the pointer; pinching with both hands resizes it.

The tracking part was the easy bit. The hard part was making it feel reliable: no jitter, no accidental triggers, and nothing that tires your arm. I wrote a jitter filter and a gesture state machine that turn Apple Vision’s hand landmarks into pointer and window actions through the macOS Accessibility API. It all runs on your Mac: video is never saved or sent, and the app has no accounts, analytics or network access.

8
Gestures, from palm-to-start to two-handed resize
0
Frames saved or sent: tracking runs fully on-device

Python · PyObjC · Apple Vision · AVFoundation · Accessibility API

Apple Silicon Intel View on GitHub ↗ macOS 12+ · ⌃⌥⌘G turns it on and off