Biography
I recently defended my PhD on the neural machine translation of user-generated content (like social media comments) at Inria Paris and Sorbonne Université. Before that, I completed an Engineering Master’s (Diplôme d’Ingénieur) in Applied Mathematics and Computer Science from Centrale Nantes. Beyond research, I enjoy sharing knowledge through writing and public speaking, and I am passionate about exploring new languages and cultures.
Open to Opportunities: I am currently seeking an AI/NLP research scientist or engineer position in industry, either on-site in the Paris area or remotely from another location.
Nishimwe is a Rwandan name meaning ‘Thanks be to God’. It is pronounced /niːʃiːmŋé/.
Fun fact: There is another Lydia Nishimwe, who is a singer. Though we share quite a few similarities, we are not related. Feel free to check out her YouTube.
- Machine Translation
- Generative AI
- NLP for Low-resource Languages
- Multimodality
-
PhD in Computer Science, 2021-2025
Inria Paris, Sorbonne Université
-
MEng in Mathematics and Computer Science, 2017-2021
École Centrale de Nantes
-
BSc in Mathematics and Computer Science, 2014-2017
Université Grenoble Alpes
Languages
Native
Native
Advanced
Intermediate
Intermediate
Elementary
Experience
Topic: Robust Neural Machine Translation (NMT) of User-Generated Content (UGC)
Supervised by Benoît SAGOT and Rachel BAWDEN, defended on June 18, 2025
- Designed data augmentation strategies and MLM-based lexical normalization techniques to improve NMT robustness on noisy user-generated content
- Developed RoLASER, a sentence embedding model trained via knowledge distillation to align noisy and standard text representations, and extended it to RoSONAR, a custom NMT system for UGC
- Evaluated LLMs for UGC translation and proposed improved dataset-specific evaluation practices
- Trained language and translation models in a high-performance computing (HPC) environment
- Published 3 first-author papers in peer-reviewed NLP venues (2 conferences, 1 journal)
- Contributed bug fixes and new features to NLP repositories on GitHub (Fairseq, NL-Augmenter)
- Gave 15+ public presentations of my work (conferences, seminars, high school outreach programs)
- Collaborated on 2 NLP shared tasks (1 in the organization team, 1 as a submission team member)
Tech stack: Python, PyTorch (Fairseq, Hugging Face Transformers), Pandas, Scikit-Learn, SLURM
Organisation: Github/Gitlab, Trello, Zotero
Office pack: LaTex/Beamer, MS Word/PowerPoint/Excel
Topic: Inference of masked sequences
Supervised by Tanguy URVOY
- Conducted an extensive literature review on sequence models (seq2seq) decoding strategies (autoregressive, semi-autoregressive, non-autoregressive, monotonic, and non-monotonic)
- Designed and executed experimental studies on decoding algorithms for reconstructing masked sequences of router logs from Orange
Tech stack: Python, TensorFlow, Keras, Pandas, Scikit-Learn
Organisation: Gitlab, Trello, Zotero
Office pack: LaTex/Beamer
Implemented the integration of a third-party app (Opencast, a Learning Management System) with RoCamRoll (the company’s product), including troubleshooting API interactions, and ensuring smooth data flow between systems.
Tech stack: Erlang, HTTP
Topic: Functional verification of an ARM7 microprocessor
- Implemented the simulation of the microprocessor in VHDL
- Implemented new features in C and their test files in ARM
Tech stack: VHDL, C, ARM
🏆Stage d’excellence (Excellence Internship Program) - Université Grenoble Alpes🏆
Implemented the simulation of GPS logs
Tech stack: Lutin, Lustre
🏆Stage d’excellence (Excellence Internship Program) - Université Grenoble Alpes🏆
