Curriculum vitae
Satish Bajagain · Lalitpur, Nepal
satishbajagain@gmail.com · Portfolio · Learning notebook · GitHub
Research interests
I am interested in learning from brain signals, particularly generalization across individuals in neural decoding. My EEG sleep-stage study explored this question, and I would like to pursue related work in temporal modeling and brain–computer interfaces (BCIs).
Education
B.Sc. Computer Science and Information Technology
Tribhuvan University — Kathford International College of Engineering and Management
2022–2026 · Kathmandu, Nepal
- 85.14% aggregate, Distinction; above 80% in all eight semesters. Tribhuvan University awards Distinction at 80% or above in aggregate.
- Relevant coursework: Artificial Intelligence; Design and Analysis of Algorithms; Mathematics I & II; Statistics I & II; Numerical Method; Simulation and Modeling; Data Warehousing and Data Mining; Theory of Computation.
Scholarships and academic honors
Merit (Jehendar) Scholarship — full undergraduate tuition
Kathford International College, under Tribhuvan University guidelines · 2022–2026
- My tuition was fully covered throughout all eight semesters. I earned the first award through the entrance exam and the next seven through my semester results.
- Ranked first among male students in my Kathford B.Sc. CSIT cohort in the entrance exam and semesters I–VII.
- Received academic book awards for seven consecutive semesters and Kathford’s recognition as the best-performing B.Sc. CSIT student for the second year. Certificate summary and scholarship policy
Research experience
Generalization across people in EEG sleep-stage classification
Independent research · 2026 · Internal final evaluation completed; publication pending
Python, PyTorch, NumPy, SciPy, scikit-learn · Study overview
- Tested whether neighboring EEG segments help predict sleep stages for people not seen during training, compared with a model using only the current segment, on Sleep-EDF Expanded (Sleep Cassette).
- Implemented signal preprocessing, 30-second segmentation and five-stage label mapping; kept each person’s recordings together throughout development and final evaluation to prevent subject leakage.
- Compared spectral-feature logistic-regression baselines, a compact convolutional network, matched current-segment and temporal-context heads, and temporal decoding controls.
- Fixed the methods before evaluating held-out people; analyzed person-averaged macro-F1, per-stage errors, bootstrap uncertainty and performance around sleep-stage transitions.
- Completed internal evaluation within one source cohort, including offline models using future context. Publication, external-cohort validation and deployment remain pending.
Selected implementation and educational work
GPT from scratch — character-level decoder transformer
Completed learning implementation · 2026 · PyTorch · Notebook and code
- Implemented character tokenization, token/position embeddings, causal multi-head attention, residual decoder blocks, feed-forward layers and layer normalization, with AdamW training and autoregressive generation.
- Completed autograd, makemore and GPT study following Andrej Karpathy’s Zero to Hero, with annotated code and handwritten derivations.
Neural-network learning collection and interactive explanations
Learning collection · Interactive models
- Published seven annotated notebooks, six focused notes, an EEG essay, eight concept lessons and three interactive models: backpropagation, a leaky integrate-and-fire neuron and EEG frequency analysis.
Professional experience
HOPMEDS HEALTH TECH PRIVATE LIMITED
Full-Stack and Machine Learning Engineer
May 2026–present · Lalitpur, Nepal
- Build radiology software, including study worklists, image-viewing interfaces and the backend services that connect them.
- Work with PACS (picture archiving and communication systems), DICOM image storage and retrieval, Orthanc and OHIF Viewer.
- Develop medical-imaging preprocessing, training and evaluation pipelines; coordinate model experiments with the software team.
- Build Python/FastAPI services to make model inference available to the application.
- Contribute to PharmHops pharmacy-management software across frontend, backend, databases and deployment, including Node.js, Express and PostgreSQL.
Niyalo Creatives Pvt. Ltd.
Full-Stack Web Developer
October 2025–April 2026 · Kathmandu, Nepal
- Developed client-facing applications across frontend, backend, REST APIs, databases and deployment workflows using JavaScript/TypeScript.
- Contributed software for UR Field Lab, a climate and disaster-risk research event, and e-commerce infrastructure. This role included my eighth-semester academic internship.
Avana Game Studio LLC
Game Developer
May 2023–February 2026 · Remote, USA-based organization
- Developed algorithmic systems in C and Python for pitch detection, MIDI comparison, pattern detection, similarity analysis and real-time audio processing.
- Built player-performance data processing and visualization, gameplay physics and 3D simulation systems.
Game Development Intern
October 2022–April 2023 · Remote
- Worked on a guitar-learning game with pitch detection and real-time audio processing, and a turn-based multiplayer ship-combat simulator.
Teaching and mentoring
Git and GitHub workshop — instructor and organizer
Kathford International College · 2024
- Taught students to manage repositories, branch and merge code, and work together through pull requests.
Frontend web development workshop — co-instructor and co-organizer
Kathford International College · 2024
- Co-organized and taught a practical introduction to frontend web development for undergraduate developers.
Technical skills
- Programming: Python, C/C++, JavaScript/TypeScript; additional experience with C#, Go.
- Machine learning and analysis: PyTorch, scikit-learn, NumPy, SciPy, pandas, Matplotlib; neural networks, convolutional models, decoder transformers, model training and evaluation.
- Signals and research methods: EEG preprocessing, spectral features, physiological time series, temporal modeling, person-disjoint evaluation, per-stage analysis and bootstrap uncertainty; audio pitch detection and similarity analysis.
- Medical imaging and systems: DICOM, PACS workflows, Orthanc, OHIF; FastAPI, Node.js, Express, PostgreSQL, Docker, GNU/Linux, CUDA/GPU workflows.
- Research and development tools: Git/GitHub, Jupyter, LaTeX and GDB. Currently studying spiking neural networks and biologically inspired computation.