At the Space and Astrophysics Research Lab, I prepared SDSS quasar spectra for deep-learning applications, processing more than 5,100 spectra and visually classifying more than 1,500. I documented the preprocessing work; model training had not begun during my placement.
My final-year project at the Institute of Space Technology applies deep learning to celestial-object classification and redshift estimation from SDSS spectra. I’m responsible for the spectral-model stream and am reproducing FNet II as the project baseline.
At PINSTECH’s LINAC facility, I mapped quadrupole and bend-magnet fields, analysed beam-spot images in MATLAB, and wrote Siemens PLC ladder logic for temperature conditioning. The measured magnet fields agreed with design parameters, and the magnets were cleared for beamline integration.
I also built a client’s scheduled n8n and Gemini pipeline to filter and rewrite trending news before posting it to Telegram. Alongside these projects, I’m developing a transformer-based next-word prediction project and have tutored students in school and engineering-diploma subjects.

