AI Engineer · Doctoral Researcher at Charles Darwin University
I build intelligent systems and study how people can remain informed, capable and accountable when AI enters real decisions.
Selected work · Research · Technical stack · Approach · Connect
I work where applied AI engineering meets human-centred system design. My current PhD research at Charles Darwin University (Darwin, Australia) examines meaningful human judgment, oversight, and standards-aligned controls in generative-AI workflows, with teacher education as the initial research setting.
|
Human agency · Responsible GenAI · Socio-technical evaluation What information, authority, and practical opportunity must a person retain when an AI system supports consequential decisions? |
Document AI · NLP / Retrieval · Vision · Evaluation I build experiments and systems with attention to baselines, failure modes, traceability, reliability and usable interfaces. |
A Streamlit research dashboard integrating sentiment analysis, topic discovery and semantic retrieval across Reddit discussions. |
My current doctoral research: translating responsible-AI principles into decision boundaries, governance controls and testable evidence. Research in progress, not completed findings. |
An experimental CNN-to-variational-quantum-circuit pipeline using PyTorch and PennyLane, with a classical baseline for comparison. |
Collaborative robotics work investigating model failure on robot-camera images and performance recovery after deployment-specific adaptation. |
Additional experiments, open resources & systems ↗
- Nepali NLP Resources — curated multilingual and low-resource NLP reference index.
- Non-Markovian ML trajectories — quantum dynamics research-code experiments.
- Reinforcement-Learning Benchmark Suite — documented training and evaluation comparisons.
- Production AI Reliability — documented evaluation approach for document-intelligence pipelines; confidential work is not published as source code.
- All public repositories — explore the wider collection.
Core languages, research and ML
Tools used across projects
The recurring principle across my work: a system is only as trustworthy as the evidence around its decisions. Model outputs matter, but so do inputs, failure analysis, review boundaries, and what happens in deployment.
I maintain public technical reports and preprints alongside my doctoral research. The records below are non-peer-reviewed technical outputs, not journal papers.
| Selected technical output | Record |
|---|---|
| FieldTrace-OCR: reliability-oriented evaluation for invoice-parsing pipelines | DOI ↗ |
| A Lightweight Multilingual Semantic Search Pipeline | DOI ↗ |
| Temporal Context and Momentum-Aware Reward Shaping in Reinforcement Learning | DOI ↗ |
| A Retrieval-Augmented Financial Advisory Prototype with Evidence-Gated Response Generation | DOI ↗ |
ORCID research record ↗ · Research programme ↗ · Books & writing ↗
Background, wider interests & selected learning
- Education: PhD, Charles Darwin University (in progress); Master of Information Technology (Artificial Intelligence), Macquarie University; BSc (Hons) Computing, First Class Honours.
- Experience: AI/ML R&D at Truuth; earlier technology leadership as CTO at Picpoint Nepal.
- Selected learning: Harvard CS50x, AI governance, design thinking and sustainability leadership.
- Beyond a single discipline: machine learning, human-computer interaction, robotics, technology leadership, digital well-being and nonfiction writing.
For a fuller and more current professional record, see my portfolio.
Research collaboration · AI engineering · Intelligent systems
Portfolio · LinkedIn · ORCID · More projects
Build what can be inspected. Measure before claiming. Keep human judgment explicit.