What shaped
the answer?
Relevance tracing for transformer and Mixture-of-Experts models.
RESEARCH NOTES / 02
Follow the
evidence.
I’m interested in how AI systems represent, retrieve, and explain information—and in making the reasoning path easier to inspect.
INQUIRY
→ METHOD
→ SYSTEMR
↘
→ METHOD
→ SYSTEMR
↘
AREAS OF INTEREST
What shaped
Where did it
come from?
Source attribution, retrieval workflows, and evidence-grounded synthesis.
Can it work
in a system?
Applied AI engineering with Python, PyTorch, and research tooling.
PUBLICATION RECORD
Work with
a paper trail.
Publication states below follow the available profile record. Open the publisher page for its current metadata.
PUBLISHEDINDIC NLP
SafeSpeech
Indic hate-speech mitigation covering detection, word-level intensity estimation, and semantic-preserving rewriting.
PAPER RECORDEXPLAINABILITY
DLBacktrace
Research record associated with relevance tracing and model explainability.
Following a related question? I’d be glad to compare notes.
Start a conversation