# Theodoros Psallidas

Senior Software Engineer at ProxyFoods building frontier agentic platforms with LangGraph. PhD candidate in Computer Science, University of Thessaly.

Canonical: https://theopsall.github.io/

## Experience

### Senior Software Engineer, [ProxyFoods](https://proxyfoods.ai)
Remote · Greece · Feb 2025 - Present

- Formulation engine for nutrition, cost, sensory profiling, and mass-balance validation.
- Agentic recipe generation with LangGraph, RAG, and human-in-the-loop ingredient matching.
- Cross-service authorization with OpenFGA and shared auth, tracing, logging, and exception libraries.

Stack: React 19, TypeScript, FastAPI, Python, LangGraph, OpenAI, Langfuse, OpenFGA, SQLAlchemy, Azure SQL, Docker, Azure, OpenTelemetry

### ML Engineer & Software Developer, [MyTripMyWay](https://mytripmyway.uth.gr/)
Greece · Jun 2025 - Mar 2026

- Persona-based recommendations with six interpretable travel personas and itinerary feasibility modeling.
- Full-stack POI aggregation, booking and payment flows, and interactive map-based itineraries.

Stack: Next.js, React, MobX, Tailwind CSS, Leaflet, FastAPI, SQLAlchemy, MariaDB, Docker, Google Maps, Stripe

### Full Stack Software Engineer, [Behavioral Signals](https://behavioralsignals.com)
Remote · Greece · Mar 2024 - Feb 2025

- Delivered agent matching, analytics, monitoring, and scalability improvements for real-time routing.
- Designed, built, and owned the audio deepfake detection application end-to-end.

Stack: ReactJS, FastAPI, Django, Docker, Ansible, Azure, AWS, PostgreSQL, Milvus, MLflow

### Machine Learning Engineer, [NCSR Demokritos](https://www.demokritos.gr)
Remote · Greece · Dec 2022 - Feb 2024

- Built music copyright monitoring and inference pipelines for IoT and Raspberry Pi fleets.
- Migrated models to deep learning and FAISS vector search to improve detection latency and accuracy.

Stack: ReactJS, FastAPI, PyTorch, FAISS, PostgreSQL, Docker, Balena, AWS, MLflow

### Lead Software Engineer, [Optechain](https://optechain.com)
On-site · Argyroupoli, Greece · Oct 2020 - Feb 2024

- Led frontend and backend development for digital signage and EV-charging stations.
- Ran the Azure and Docker stack and managed device-fleet updates for safer rollouts.

Stack: React, Angular, Node.js, C#, Azure, Docker, MSSQL, Balena, Mender

### Machine Learning Engineer, [NCSR Demokritos](https://www.demokritos.gr)
Remote · Greece · Dec 2019 - Jul 2022

- Built video summarization and multimodal analysis models for the Enorasi project.
- Improved summarization through model architecture and feature-fusion experiments.
- Extended benchmark labeling to reduce evaluation variance.

Stack: PyTorch, scikit-learn, Python, Computer Vision, Multimodal ML

## Publications

- 2024. [A multimodal dataset for electric guitar playing technique recognition](https://doi.org/10.1016/j.dib.2023.109842). A. Mitsou et al., Data in Brief
- 2023. [Video summarization based on feature fusion and data augmentation](https://doi.org/10.3390/computers12090186). T. Psallidas, E. Spyrou, Computers
- 2022. [Summarization of user-generated videos fusing handcrafted and deep audiovisual features](https://doi.org/10.1109/SMAP56125.2022.9941864). T. Psallidas, E. Spyrou, S. J. Perantonis, SMAP
- 2022. [ENORASI assistive computer vision-based system for the visually impaired: a user evaluation study](https://doi.org/10.1145/3529190.3534784). A. Mitsou et al., PETRA
- 2022. [Multimodal video summarization based on fuzzy similarity features](https://doi.org/10.1109/IVMSP54334.2022.9816266). T. Psallidas, M. D. Vasilakakis, E. Spyrou, D. K. Iakovidis, IEEE IVMSP
- 2021. [Multimodal summarization of user-generated videos](https://doi.org/10.3390/app11115260). T. Psallidas et al., Applied Sciences
- 2020. [ARCHEO: a dataset for sound event detection in areas of touristic interest](https://doi.org/10.1109/SMAP49528.2020.9248467). T. Psallidas et al., SMAP
- 2020. [Dimensionality reduction and attention mechanisms for extracting affective state from sound spectrograms](https://doi.org/10.1007/978-3-030-66125-0_3). G. Pikramenos et al., ICPRAM

## Education

- **PhD Candidate in Computer Science**, University of Thessaly (2021 - Present)
- **MSc in Data Science**, University of Peloponnese (2019 - 2021). Thesis: Multimodal summarization of user-generated videos from wearable cameras
- **BSc in Computer Science**, University of Thessaly (2013 - 2018)

### Certifications

- [Back End Development and APIs](https://freecodecamp.org/certification/theopsall/back-end-development-and-apis), freeCodeCamp (2022)
- CCNA Routing and Switching course, University of Thessaly (2016 - 2017)

## Open source

- [Video-Summarization](https://github.com/theopsall/Video-Summarization): Code and dataset for my MSc thesis on multimodal summarization of user-generated videos from wearable cameras. (Jupyter Notebook)
- [multiSmote](https://github.com/theopsall/multiSmote): A multi-label approach to the SMOTE oversampling algorithm. (Python)
- [deep_video_extraction](https://github.com/theopsall/deep_video_extraction): Deep visual and audio feature extraction from video with pre-trained models. (Python)
- [video_annotator](https://github.com/theopsall/video_annotator): Web app for annotating video shots, used to build the thesis ground truth. (Python)

## Contact

- [Email](mailto:theopsall@gmail.com): theopsall@gmail.com
- [GitHub](https://github.com/theopsall): github.com/theopsall
- [LinkedIn](https://www.linkedin.com/in/tpsallidas): linkedin.com/in/tpsallidas
- [Scholar](https://scholar.google.com/citations?user=478yYkIAAAAJ): Google Scholar profile
- [Work](https://proxyfoods.ai): proxyfoods.ai