I started working in Natural Language Processing, but after 15+ years, I have worked in Vision, Speech, and various exotic datatypes (auction data, log data, time series, ..), in addition to natural language. My ‘research taste’ is in the state-of-the-art that enables magical product experiences.
Current Research Interests
- Low resource Modeling
- Alignment & Explainability
- Multilingual modeling
- Multimodal modeling
- Instruction-finetuning
- Datasets
- Reasoning
- Tool invocation
- Discovery
- Reliability
- Evaluation
- Code
- Log files
- Organic Compounds (eg. SMILE)
- Time Series
- . . .
- Encoding Linguistic Knowledge
- “World knowledge”
- Knowledge Graphs & Ontologies
I am also interested in “exploring the automation spectrum”, particularly where humans and machines work as one unit. The displacement of humans will become inevitable if we keep giving our data to big corporations who build models behind walled gardens on the backs of our data. If I were to give a one-sentence description of my current/future work thesis, it would be “active strategies to enable progress and prevent displacement of humans by AIs.”