Delft · TU Delft
Andrej Bošnjak
I build embedded systems, quantitative models, and useful software.
About
I am based in Delft. In August 2026 I am joining Montana Aerospace as an engineering intern.
In September 2026 I am joining AeroDelft as an electronics engineer on the Sling 4 hydrogen retrofit, working on high and low power electronics and aircraft control management.
- Education
- BSc Aerospace Engineering, TU Delft, 2025-present IB Diploma, 42 / 45, VIS, 2023-25
- Focus
- Data and modelling for decision-making under uncertainty
- Tools
- Python, FastAPI, PostgreSQL, Docker, React, Supabase, Vite, CATIA, Fusion 360
- Languages
- English C2, German C1, Serbian native, Spanish B1
Projects
-
lidar_tracking_gimbal
2026-Real-time 3D object tracker: an Arducam ToF camera on a Raspberry Pi detects targets and sends positions to an ESP32-driven pan/tilt gimbal. Multi-object Kalman tracking and dual-axis PID run over UART.
-
imc_prosperity_4
2026Algorithmic trading system for IMC Prosperity 4. Top 10 in the Netherlands in qualifiers and 7th after the first finals round; strategies cover market making, relative value, and arbitrage with explicit inventory controls.
-
hackdelft_lumen
2026Privacy-first financial intelligence platform for bank statement exports, with ingestion, deduplication, analytics, subscription detection, AI-assisted querying, and PDF reporting. ML categorisation pipeline using e5-large embeddings and a dense neural classifier; 95.7% top-1 accuracy on real data.
-
delft_dibs
2026-Co-founded a housing platform for Delft's competitive student rental market. Scrapes and normalises listings from 11 rental sites through a FastAPI/PostgreSQL backend. Core feature is an OLS relative-value model assigning a live deal score to each listing.
-
audio_spectrogram_visualizer
2026Interactive Fourier audio explorer with WAV playback, live microphone input, configurable FFT analysis, band-pass filtering, and MP4/GIF spectrogram exports.
-
breadwinner
2026AI-assisted desktop workspace for hardware debugging. Connects to serial devices and brings serial output, datasheet pinouts, project logs, and controlled diagnostic scripts into one project context.
-
bitcoin_pow_simulation
2024-25Stochastic model of Bitcoin's proof-of-work mechanism: mining difficulty adjustment, hash rate dynamics, and reward distribution under varying network participation.
Reading
Books
- Think and Grow Rich - Napoleon Hill
- Atomic Habits - James Clear
- Greenlights - Matthew McConaughey
- Divine Comedy - Dante Alighieri
- Crime and Punishment - Fyodor Dostoevsky
- White Nights - Fyodor Dostoevsky
- 1984 - George Orwell
- Frankenstein - Mary Shelley
- The Picture of Dorian Gray - Oscar Wilde
- A Moveable Feast - Ernest Hemingway
- The Old Man and the Sea - Ernest Hemingway
- Animal Farm - George Orwell
Papers
- Attention Is All You Need - Vaswani et al., 2017
- Training Compute-Optimal Large Language Models - Hoffmann et al., 2022
- In-context Learning and Induction Heads - Olsson et al., 2022
- Scaling Monosemanticity - Templeton et al., 2024
Contact
Email: [email protected]
GitHub: github.com/andrejbosnjakk
LinkedIn: linkedin.com/andrej-bosnjak