Aug 31, 2026 Published Harness Engineering 101, a blog series on building LLM agents and harnesses from first principles. Sixteen chapters, one JSON array.
Aug 19, 2026 Paper accepted to Findings of EMNLP 2026 (first author): “Confident but Wrong: A Constrained Decoding Diagnostic for Low-Resource Automatic Post-Editing”.
Jul 31, 2025 Started as an external volunteer researcher with the University of Moratuwa, working on Google-funded research on Automatic Post-Editing for low-resource languages (Sinhala, Tamil).
Feb 28, 2025 Presented a technical deep-dive on blockchain architecture, consensus algorithms, and distributed ledger mechanisms at WSO2 Technology Conference 2025. The recording is up, slides and all, on YouTube.
Jan 14, 2025 Released lightweight code retrieval models (22M/33M parameters) on Hugging Face: L6 and L12, achieving 97% Recall@10 for domain-specific code search.