Abhilash Pal

Abhilash Pal

An adept researcher in NLP and deep learning, with a track record of leveraging advanced models for diverse applications in industry and academia.

Curriculum Vitae

Welcome to my personal website, feel free to look around. Do reach out if you want to connect with me. The best way to connect with me is via email.

Abhilash Pal

Research Interest

  • Machine Learning
  • Natural Language Processing
  • Computer Vision
  • Robustness
  • Interpretability

Work Experience

WerkStudent in NLP/Robust Deep Learning, iABG, Munich, Germany June’23 – Present

  • Utilizing LLMs for code generation and topic modeling.
  • Building and maintaining a coding assistant and extending it to a general chatbot.
  • RAG project for extracting supporting evidence for user queries.

WerkStudent: Applied Scientist in Speech Recognition, Microsoft, Munich, Germany Oct’21 – Mar’23

  • Working on Automatic Speech Recognition and Sign Language Recognition.
  • Adapting internal testing tools and providing effective remedies.
  • Sign Language Detection and Translation tasks.

IDP Student, Convaise UG, Munich, Germany Apr’21 – Sept’21

  • Working on Text Generation using GPT2, Prophetnet, T5, and Pegasus.
  • Multilingual Question Generation and Paraphrasing.

NLP Intern, Citrusberry Business Solutions, Mumbai, India Dec’19 – Aug’20

  • Trained GPT2 for Product Description Generation.
  • Scraped Websites and cleaned textual content for Dataset Generation.

Summer Research Intern, Indian Statistical Institute, Kolkata, India May’19 – July’19

  • Built an OCR for reading historical documents in Bengali script.
  • Scraped Bengali News websites for dataset.

Computer Skills

  • Programming: C, C++, Python, Javascript, Bash
  • DL Tools: Pytorch, Scikit-learn, Tensorflow, Huggingface, vLLM, Langchain, BERTopic
  • OS: Linux, Windows
  • Cloud: Azure, AWS Sagemaker
  • Other Tools: Docker, Slurm

Education

  • MSc. In Informatik, Technische Universität München(TUM), Germany
    • Cumulative Grade - 2.1
    • Final Thesis: “Interpretable and Faithful Topic Modeling” - 1.7
  • Bachelors in Technology - Computer Science, SRM Institute of Science and Technology, India
    • CGPA - 9.1 out of 10
    • Final Project: MuseNet - Music Generation using LSTMs.

Conference Publications

  • “Clean Text and Full-Body Transformer: Microsoft’s Submission to the WMT22 Shared Task on Sign Language Translation” in EMNLP 2022.
  • “On The Importance of Signer Overlap For Sign Language Detection” (available in arxiv).

Journal Publications

  • “Musenet: Music Generation using Abstractive and Generative Methods”, in International Journal of Innovative Technology and Exploring Engineering, 2020.
  • “ConciseNet: An End to End Abstractive Model for Topic Generation”, in International Journal of Engineering Applied Science and Technology, 2019.

Achievements

  • Submission to “WMT shared task on sign language translation 2022” was ranked number #1 in terms of BLEU score.
  • First Prize at IET Hackathon 2019, SRM University KTR.