Hi, I'm Aflah, a research software engineer at MPI-SWS. I specialize in backend engineering, NLP, and deep learning, with a keen interest in applying NLP for social good. I am also interested in high speed inference, large scale distributed training and squeezing out performance on limited hardware. Currently, my research involves evaluating latent knowledge and memorization in large language models. I have previously worked on hate speech detection, and leveraging NLP to address various societal challenges as a part of undegrad thesis and related research work.
Open to researcher/research engineer/backend engineer roles
- Trying to understand PiPPy and Pipeline Parallelism API in PyTorch
- Paper Reading & Discussion: LoRAMoE: Alleviate World Know. Forgetting in LLMs via MoE-Style Plugin
- Paper Reading & Discussion: TorchTitan: One-stop PyTorch native solution for production ready LLM...
- Finding the Best Number Tokenization Method in LLMs
- Paper Reading Discussion: Lookback Lens: Detecting and Mitigating Contextual Hallucinations
