Our Private Tutors

Education

• Universitas Indonesia

Expertise in Subject:

• Information Technology (Cambridge AS & A Level)
• Mathematics (Cambridge AS & A Level)
• Further Mathematics (Cambridge AS & A Level)
• Physics (Cambridge AS & A Level)

Work Experience

• International Collaboration with Prof Derry Wijaya from Boston University – 1 | Research Collaborator Dec 2022 – Jul 2023
  • Recognized the urgent problem of the absence of replicable benchmarks for low-resource Neural Machine Translation
  (NMT). Highlighting that existing benchmarks in NMT often consist of billions of parameters, well outside the reach of
  many researchers.
  • Led the Development of Replicable Low-Resource NMT Benchmarks, creating and advocating for the use of accessible yet robust evaluation tools for the Indonesian research community. Achieved by establishing a replicable NMT
  benchmark for four low-resource local languages in Indonesia, providing a template for future works.
• International Collaboration with Prof Derry Wijaya from Boston University – 2 | Research Collaborator Jul 2023 – Dec 2023
  • Identified that LLMs do not perform well for both high-resource and low-resource translation tasks. Drawing attention that
  the State-of-the-art in NMT is held by a supervised NMT model, comparatively smaller in parameter size and orders of
  magnitude smaller in GPU training time than LLMs such as Llama-2.
  • Contributed to the field of NMT by showing the upper-bound performance of LLaMa-2 models on translation from
  English to an artificial language we created. Our findings reveal that their performance falls short of supervised NMT
  models, suggesting that scaling up LLMs does not efficiently help translation performance.
  • Investigated multiple variables such as information ordering and word-type availability during prompting. Not having noun
  translations and shuffling the word-level translations in the prompt degrades the model most. Meanwhile, masking other
  word types does not significantly affect the model.
• International Collaboration with Prof Derry Wijaya from Boston University – 3 | Research Collaborator Jul 2023 – Dec 2023
  • Recognized that LLMs are multilingual, impacting their translation performance. Exploring the diverse attributes of languages and their influence on LLMs’ translation performance.
  • Contributed to the future development of NMT systems by identifying which linguistic features and languages
  have a high impact on the model’s translation performance, aiming to improve the performance of low-resource
  language translation by strategically utilizing and selecting higher-resource languages.
• Kata.ai | AI Research Intern Feb 2022 – Jul 2022
  • Developed various training curricula not tested before and evaluated their impact on BERT’s performance for the GLUE
  benchmark.
  • Conducted in-depth evaluations of multiple progressive stacking techniques to enhance the efficiency of BERT pre-training.
  • Created a pipeline for pre-training BERT models utilizing both training curriculum and progressive stacking which potentially increases BERT’s pre-training convergence rate by 4% while reducing compute requirements by 12%.
• University of Indonesia | Teaching Assistant for Design and Analysis of Algorithms Course Aug 2021 – Jan 2022
  • Tutoring students on analyzing the time complexity, memory complexity, and the correctness of various algorithms.
  • Guiding some students in writing papers for the course’s final assignment.
• University of Indonesia | Teaching Assistant for Database Course Feb 2021 – Jul 2021
  • Tutoring students on Database Modeling and PostgreSQL basics.
  • Evaluating and providing personalized feedback to students on their assignments