Shantam Raj
Curious · Driven · Poet
I write code when the sun’s high,
and poetry when the moon’s nigh.
about
My work is interdisciplinary - NLP and LLMs at the core, with a slow, deliberate turn toward policy. I’m drawn to AI in education and to econometric methods that surface effects at scale.
I am looking for ambitious collaborations for social good.
now
Almost there: research on human preferences, narratives, LLM alignment and causal NLP with Prof. Maria Antoniak, University of Colorado Boulder and Prof. Elliott Ash, ETH Zurich.
Soon: Studying the cross-transfer of skills in LLMs with Prof. Mrinmaya Sachan and Dr. Mubashara Akhtar.
experience & affiliations
-
ETH AI Center Affiliated Researcher
-
University of Zurich Research Assistant
-
SUREAL AI Lab Researcher
selected publications
Public talks
selected work (GitHub has the latest)
- Lucan lucan.earth
- Narrative Reward Modelling Reward models for LLM alignmentcoming soon
- Lobbying and Nature biodiversity & corporate influence coming soon
- Cross-domain skill transfer in LLMs skill transfer in LLMscoming soon
- Digital Twin — Satellite Pipeline
- trama Trope detection with LLaMA 2
- oh-my-slurm-eval Running LLM evals on Slurm clusters
- Algowritten Project with Mozilla on learning the biases in LLMs
- Covid Viz Data visualisation of the pandemic
- Gifc Command-line GIF tooling
- Template Conformance
- Editorial Agent
- active or passive
- Beej Retired