Skip to content
Darwin, NT · AustraliaDoctoral Researcher · Charles Darwin UniversityDRW --:--:-- · UTC+09:30

Applied AI · Human-Centred Systems · Responsible Generative AI

Shaurav Khadka

I build reliable AI systems and research how intelligent tools can augment human judgement without eroding human agency.

My work spans production AI reliability, retrieval, computer vision and robotics, and technical governance—drawing on industry AI/ML R&D, more than three years of CTO-level leadership, and doctoral research at Charles Darwin University.

View applied systemsMeasured resultsRésumé
Observatory AtlasGitHubLinkedIn

Measured Highlights

Start with what changed.

Three benchmarks across deployment adaptation, reinforcement learning, and retrieval.

Deployment-specific Sim2Real adaptation

2.38% → 95.24%

Baseline2.38%
Adapted95.24%

Robot-image accuracy after deployment-specific Sim2Real adaptation

The model looked strong on curated data and degraded sharply on robot-camera images. The recovery came from treating domain shift as a deployment problem, not a footnote.

Why it matters: the adaptation restored useful robot-camera performance under changed lighting, viewpoint, scale, and background conditions.

Inspect case study View code
Robot-camera deployment prediction comparison after Sim2Real adaptation

Robot-camera predictions · published team-level result

Evaluation context
Baseline
2.38% before deployment-specific adaptation.
Measured
95.24% robot-image accuracy after targeted collection, augmentation, and fine-tuning.
Conditions
Robot-camera inputs with lighting, viewpoint, scale, and background differences.
Attribution
Collaborative team-level result with exported notebook figures.

Supporting benchmark 02

300 → 1,925

AirRaid PPO mean reward after temporal observation changes

Observation design materially changed what the policy could learn. Frame skipping and frame stacking improved the benchmark result without pretending algorithm choice was the only lever.

Why it matters: the result shows that observation design can materially change what a policy learns before the algorithm itself is replaced.

Inspect case study

Supporting benchmark 03

P@5 = 0.68 · R@5 = 0.68

RedditPulse semantic retrieval quality

The retrieval layer was measured before generation was treated as useful. That matters because grounded insight quality depends on which sources the system surfaces first.

Why it matters: downstream summaries are only as useful as the source material retrieved before generation begins.

Inspect case study View code Live demo

Applied Systems

Selected applied systems.

Production reliability, temporal learning, semantic retrieval, conversational AI, and robotics—with each system linked to an inspectable case study.

Prior industry workflowTRUUTH · Former AI/ML R&D Internship

Production AI Reliability and Document Intelligence

ProblemDocument intelligence can fail long before or after OCR. Real reliability depends on the complete path from ingestion to extraction, transformation, validation, and review.

ContributionBuilt repeatable evaluation workflows across OCR configurations, mappings, confidence scores, error codes, and reruns while preserving traceability and review boundaries.

Inspect case study

01

OCR

02

Map

03

Validate

04

Trace

Shared here: sanitised workflow record. Confidential operational data and internal metrics are excluded.

Selected systems / case studies

Browse selected work by problem.

Five selected builds spanning temporal learning, conversational AI, retrieval, and vision/robotics.

Temporal Graph LearningResearch build

Temporal GNN for Blockchain Fraud Detection

Fraud is relational and time-dependent. Static tabular features can miss how transactions evolve across a network.

System tracet0 → t1 → t2
Open case study

Research Profile

Research

Doctoral research on responsible Generative AI, human agency, and standards-aligned system design.

Teacher education is the initial empirical domain. The broader question is how responsible-AI principles and standards become testable controls, evaluation criteria, governance mechanisms, and human-oversight boundaries in real deployments.

Discuss research or collaboration
Current doctoral research

Human Agency and Responsible Generative AI

Designing and evaluating a standards-aligned socio-technical reference framework for responsible Generative AI, with teacher education and professional learning as the initial empirical domain. The research connects technical controls with governance, human oversight, privacy and data governance, transparency, AI literacy, and institutional decision-making.

Doctoral research program

Questions I am building toward

  • How can responsible-AI principles and standards be translated into testable technical and organisational controls?
  • Which decisions should remain meaningfully human, and how should oversight and escalation boundaries be designed?
  • How should privacy, transparency, traceability, AI literacy, and system effectiveness be evaluated together rather than in isolation?
Responsible GenAI research routeProduction reliability foundation

Applied research foundation

Reliable AI Systems and Production Evaluation

Evaluation of AI pipelines where traceability, robustness, confidence handling, validation dependencies, regression risk, latency, cost, and human-review boundaries matter alongside headline accuracy.

Active research area

Retrieval, Grounding and Knowledge-Centred AI

Semantic retrieval, RAG, multilingual discovery, and evidence-gated generation — with emphasis on whether the system retrieves the right evidence before generated language is treated as useful.

Applied systems direction

Human-Centred AI, Decision Support and Digital Transformation

A systems direction connecting AI engineering, organisational workflows, governance, decision support, and technology adoption. The focus is not automation for its own sake, but designing digital systems that improve decisions, preserve meaningful human control, and produce measurable operational value.

Research mapResearch architectureThree foundation groupsOpen map

The program connects governance and standards to technical evaluation rather than treating them as separate conversations. Existing work in production reliability, retrieval, temporal modelling, and deployment adaptation provides the applied base.

Responsible GenAI and governance

Current doctoral layer: operationalising responsibility into system and organisational design.

  • Standards alignment
  • Human agency
  • Human oversight
  • Risk controls
  • Privacy and data governance
  • Transparency
  • Traceability
  • AI literacy

Evaluation and deployment

Applied base built through production-oriented AI R&D and benchmarked systems work.

  • Experiment design
  • Baseline comparison
  • Error analysis
  • Confidence analysis
  • Regression testing
  • Domain shift
  • Human-review boundaries
  • Reproducibility

Systems and leadership

Engineering and organisational experience that supports socio-technical research rather than model-only analysis.

  • AI/ML engineering
  • Data workflows
  • APIs
  • Digital systems
  • Technology strategy
  • Stakeholder coordination
  • Operational workflows
  • Decision-ready reporting

Decision systems and digital transformation

Bridging technical capability with organisational adoption, workflow design, decision quality, and measurable operational outcomes.

  • Human-centred AI
  • Decision support
  • Digital transformation
  • Workflow redesign
  • Automation assessment
  • Technology adoption
  • Data-informed operations
  • Socio-technical systems

Books / Long-Form Work

Books & Long-Form Work

Two nonfiction books on human agency, behaviour, and technology, plus selected illustration and editorial work.

Independent nonfiction · human agency · technology · behaviourOpen library
THE ARCHITECTURE OF OTHERWISE
The Architecture of Otherwise cover

Narrative nonfiction · philosophical psychology

The Architecture of Otherwise

Why We Know Better, Act Differently, and Change Under Pressure

A narrative nonfiction inquiry into the gap between knowing and doing: how habit, attention, stress, environment, social pressure, technology, and competing incentives shape behaviour and human agency.

First Edition151 pagesKindlePaperback
  • Narrative inquiry into the gap between knowledge, intention, and action
  • Connects philosophy, psychology, behaviour, context, and experimental reasoning
  • Centres human agency without reducing change to willpower alone
KindlePaperbackGoodreads
THE DIGITAL EQUILIBRIUM
The Digital Equilibrium cover

Human agency · designed digital systems

The Digital Equilibrium

Reclaiming Attention, Agency, and Well-Being in a Designed World

A substantially expanded second edition on protecting attention, judgement, privacy, relationships, autonomy, and human capability inside increasingly designed digital environments.

Second Edition150 pagesKindlePaperback
  • Substantially rebuilt with expanded research and a stronger theory of designed environments
  • 17 figures spanning attention, privacy, cognitive offloading, recommender systems, reliance, and responsibility
  • Practical audit plus a 30-day Digital Equilibrium reset
KindlePaperbackGoodreads
Selected creative catalogue6 illustration and editorial credits.

Illustrator · Creative contributor

Joyful Stories

Joyful Stories

AmazonGoodreads

Illustrator · Creative contributor

Joyful Stories

Mazzako Katha · Alternate edition

AmazonGoodreads

Illustrator · Creative contributor

2 in 1 Joyful, Children Stories

Combined children’s-story edition

AmazonGoodreads
Amazon authorGoodreads

Experience

Technical Experience & Leadership

Production-oriented AI R&D, more than three years of CTO-level technology leadership, and hands-on software engineering across AI, data, APIs, GIS, and operational systems.

  1. Truuth

    AI/ML Research and Development Intern

    Feb 2026 — Jun 2026

    Sydney, NSW, Australia · Hybrid

    Completed a 13-week industry AI/ML R&D major project on production document-intelligence reliability and adversarial fraud-detection evaluation. Built repeatable workflows across ingestion, OCR configuration, field mapping, transformations, validation, confidence review, reruns, and structured error analysis using Python, pandas, AWS S3/boto3, Azure Document Intelligence, and JSON; the project was awarded 83/100 (Distinction).

  2. Picpoint Nepal Pvt. Ltd.

    Chief Technology Officer

    Jun 2021 — Jun 2024

    Kathmandu, Nepal · Hybrid

    Owned technology strategy and continuous improvement across web platforms, databases, APIs, GIS/mapping inputs, data flows, export-logistics workflows, customer management, and digital operations. Translated organisational requirements into roadmaps, SOPs, implementable systems, and decision-ready recommendations while coordinating technical and non-technical stakeholders in a resource-constrained environment.

  3. Thakur International

    Junior Full Stack Developer

    Jun 2019 — May 2020

    Kathmandu, Nepal · On-site

    Developed and maintained web and mobile components using PHP, Python, and JavaScript; integrated REST/SOAP APIs, OAuth authentication, and Google Maps/geolocation workflows; and contributed to debugging, refactoring, performance analysis, and agile sprint delivery.

Foundation

Education & Research

Formal academic progression from software engineering and computing into applied AI and standards-aligned responsible Generative AI research.

Education

Charles Darwin University

Doctor of Philosophy (PhD) · Responsible Generative AI

2026 — Present · Casuarina Campus · Darwin, NT, Australia

Doctoral research supervised by Jon Mason: a standards-aligned technical reference framework for responsible Generative AI in teacher education, with focus on risk controls, human oversight, privacy and data governance, transparency, and AI literacy.

Education

Macquarie University

Master of Information Technology · Artificial Intelligence

Qualified 8 Jul 2026 · Sydney, NSW, Australia

Industry AI/ML R&D major project at Truuth: 83/100 (Distinction). Relevant study included Advanced Machine Learning, AI for Text and Vision, Data Science, AI Ethics and Law, Automated Decision Making, Knowledge, Planning and Decision Making under Uncertainty, and Advanced Topics in AI.

Education

London Metropolitan University · Islington College

BSc (Hons) Computing · First Class Honours

Awarded Mar 2021 · Kathmandu, Nepal

Final-year applied software-engineering project: an integrated trip-planning and travel-experience platform using PHP/Laravel, MySQL, HTML/CSS, and JavaScript.

About

Applied AI, systems, and responsible technology

I am an AI engineer, technology leader, and Doctoral Researcher at Charles Darwin University. My background combines AI/ML evaluation, software engineering, data workflows, more than three years of CTO-level leadership, and industry R&D.

I tend to work beyond the model itself: data quality, baselines, failure modes, validation, human review, operational constraints, and the decisions a system eventually influences.

My doctoral research focuses on responsible Generative AI, human agency, standards, and governance, while I continue to work across applied AI, retrieval, digital systems, decision support, and technical evaluation.

Research stance

Build what can be inspected.

Measure before claiming.

Treat governance as part of system design.

Keep human decision boundaries explicit.

Make the evidence trail stronger than the rhetoric.

Contact

Research, applied AI, and technical collaboration.

Based in Darwin, I am open to research collaborations, applied AI work, and technically ambitious projects where evidence and system quality matter.

Darwin, NT · Doctoral Researcher · Applied AIGitHub ↗LinkedIn ↗ORCID iD ↗