I'm a 21-year-old AI researcher and engineer focused on efficient deep learning systems, large language models, and post-training methods.
My interests span model architectures, attention mechanisms, tokenization, reinforcement learning for LLMs, reasoning models, and scalable training systems. I enjoy building ideas from research papers into working implementations and exploring new directions that push beyond current approaches.
- Efficient Transformer architectures
- Linear and state-space attention
- LLM post-training and reinforcement learning
- Knowledge distillation
- Tokenization and sequence compression
- Long-context language models
- Training infrastructure and optimization
Alongside research, I enjoy building production systems involving
- Large-scale AI pipelines
- Full-stack web applications
- Distributed training
- Model serving
- Data processing infrastructure
- Developer tooling
- PyTorch
- TensorFlow
- CUDA
- Hugging Face
- Transformers
- Deep Learning
- Computer Vision
- NLP
- Reinforcement Learning
- Python
- Node.js
- Express
- Java
- C++
- PostgreSQL
- MongoDB
- TypeScript
- JavaScript
- Vue.js
- Nuxt
- HTML
- CSS
- Tailwind CSS
- Git
- Linux
- Docker
- Jupyter
- Unity
https://github.com/mananchawla2005/gpukernels
https://github.com/mananchawla2005/implementations
https://github.com/mananchawla2005/adaptive-tokenization
- Efficient LLMs
- Model Compression
- Post-training Algorithms
- Reinforcement Learning
- Long Context
- Reasoning Models
- Retrieval
- Agent Systems
- Multimodal Learning
📧 Email: mailto:mananapeejay@gmail.com
💼 LinkedIn: https://www.linkedin.com/in/manan-chawla-a89855193/
🌐 Website: https://bymanan.com
📝 Blog: https://bymanan.com/blog
"I enjoy turning research ideas into working systems."