RESEARCH PORTFOLIO 2026 / EXPLORE THE WORK

Intelligent systems.
Real-world problems.
Evidence that matters.

Research at the intersection of agentic AI, medical intelligence, privacy-preserving learning, and scientific automation. Explore open-source systems, interact with research concepts, and examine selected publications.

Explore research ↗Try live demonstrations ↗

RESEARCH SOFTWARE & PUBLICATIONSDR. RAJESH KUMAR / BEIHANG UNIVERSITY (HANGZHOU) · UESTC YDRI LAB (HUZHOU)

RESEARCH_VIEWER / MEDICAL-AI01 / 03

Illustrative MRI annotation interface concept supplied as a visual reference; not a clinical scan or validated system output

INTERACTIVE RESEARCHMedical AI & Image AnalysisAnnotation · Segmentation · Human oversight

DESIGN REFERENCE · NOT CLINICAL DATAFIGURE 01 ↗

A BETTER WAY TO SHOW RESEARCH

Not just a publication list.
A window into how the work operates.

Each featured project connects the research question with software, reproducible methods, or an interactive explanation. Scientific claims are kept separate from illustrative demos.

01 / FEATURED RESEARCH

Selected work.

Research projects worth exploring beyond the abstract.

CURATED PROJECTS / 2021—2026

NEW / 2026 AGENTIC AI · RESEARCH INFRASTRUCTURE

ResearchLedger
Evidence-first research automation.

From falsifiable ideas to traceable results. An experimental research workflow connecting hypotheses, executed runs, evidence records, peer-review stages, and manuscript claims.

PROTOCOL LOCKCLAIM → EVIDENCESANDBOXED RUNSSCIENTIFIC AUDIT

Explore repository ↗Inspect workflow ↗

Software prototype • Current repository release v2.0.1 • Automated checks do not certify scientific correctness.

RESEARCHLEDGER / WORKFLOW● ARCHITECTURE

01 / DEFINEResearch ideaFalsifiable hypothesis

↓

02 / EXECUTEExperimentIsolated code + results

↓

03 / VERIFYEvidence ledgerClaim C### / Run R#### / Evidence E###

↓

04 / REPORTScientific outputEvidence-bound claims

EXPLORATION ≠ VALIDATIONRJ / 2026

Illustrative annotated axial brain MRI visualization; supplied reference concept, not verified model output

ILLUSTRATIVE INTERFACE REFERENCE

02 / MEDICAL INTELLIGENCE

AI-assisted medical image annotation

Explore segmentation masks, human corrections, and quantitative overlap metrics through a browser-based synthetic demo.

✚HOSPITAL A

✚HOSPITAL B

✚HOSPITAL C

▦AGGREGATOR

RAW DATA REMAINS LOCAL · MODEL UPDATES ONLY

03 / PRIVACY-PRESERVING AI

Federated learning for healthcare

Explore cross-institution training, client differences, privacy considerations, and the limits of decentralized learning.

02 / INTERACTIVE EXPLANATION

Inside the research workflow.

Explore the architecture described in the public ResearchLedger repository.

Read the protocol ↗

STAGE 01 / RESEARCH QUESTION

Define a falsifiable question.

A research idea becomes useful only when its assumptions and disconfirming conditions are explicit.

HYPOTHESIS → PROTOCOL → EVALUATION CRITERIA

This is an architectural explainer based on repository documentation, not a live execution of ResearchLedger or proof of scientific validation.

03 / INTERACTIVE LABORATORY

Try the workbenches.

Run browser experiments, inspect outputs, and learn what the models can—and cannot—show.

04 RUNNABLE · 02 RESEARCH PREVIEWS

These are educational concept demonstrations and previews, not clinical devices or independently validated scientific systems.

CS / CASE STUDIES

How the work was done.

Research problem, novelty, method, results, reproducibility and my own contribution for three flagship projects, with every figure taken from the cited paper or repository.

▤

AGENTIC AI · SOFTWARE

ResearchLedger

Evidence-first research automation: problem, design, 351-test development evidence, pre-registered validation and reproducibility status.

PROTOTYPE · v2.0.1Read case study →

◈

AGENTIC AI · arXiv 2026

AutoResearch

Execution-grounded multi-agent research workflow: 3.5% invalid citations and 79% execution success in controlled comparisons.

PREPRINT · FIRST AUTHORRead case study →

✚

MEDICAL AI · IEEE SENSORS J.

Blockchain-federated COVID-19 CT

Privacy-preserving multi-hospital CT detection with capsule networks and the public CC-19 dataset; 0.987 sensitivity.

JOURNAL · 435 CITATIONSRead case study →

04 / OPEN SCIENCE & TOOLING

Built in the open.

Code, research skills, and methods for scientists building with AI.

▤

SOURCE REPOSITORY

ResearchLedger

Evidence ledger, automated experiment workflows, prospective protocol controls, and claim-to-run traceability.

PYTHON · SCIENCE AUDITSGitHub ↗

⌘

SKILLS LIBRARY

Academic Research Skills

Claude + ChatGPT workflows for ideation, experimentation, paper review, figures, presentation design, and research writing.

SKILLS · RESEARCH AUTOMATIONExplore ↗

⚙

RELATED RESEARCH SOFTWARE

AutoResearch

Research workflow automation and experiment-grounded orchestration. Associated with a public 2026 preprint.

MULTI-AGENT RESEARCHSource ↗

◈

RESEARCH SOFTWARE

Beihang Referee

Research-paper assessment agents, structured critique workflows, and methodology-focused review utilities.

PEER REVIEW · EVIDENCERepository ↗

05 / RESEARCH IN MOTION

Watch the demonstrations.

Walkthroughs, workflows, and practical examples from the research ecosystem.

YouTube ↗

RESEARCH DEMONSTRATION

Research workflow showcase

A video supplied for the portfolio. Play directly or open on YouTube.

Watch on YouTube ↗

CLAUDE + CHATGPT RESEARCH SKILLS

How to Create a Full Research Paper with Claude Skills

Accompanying demonstration linked directly from the academic-research-skills repository.

Watch on YouTube ↗

06 / SELECTED PUBLICATIONS

Research outputs.

Selected work in medical AI, privacy-preserving systems, and research automation.

ORCID profile ↗

2026AutoResearch: An Execution-Grounded Multi-Agent Framework for Reliable Research Workflow AutomationarXiv preprint · arXiv:2607.02520PREPRINT↗

2026VisGen: Contract-Grounded Behavioral Evaluation of Canvas-Based Interactive GraphicsSubmitted to Computers & Graphics · no public DOI suppliedSUBMITTED—

2025MID-LLM: Enhancing Medical Image Diagnostics With LLMs in a Blockchain AI FrameworkIEEE Internet of Things JournalJOURNAL↗ 2025Privacy-preserving federated transfer learning for enhanced liver lesion segmentation in PET-CT imagingArtificial Intelligence in MedicineJOURNAL↗ 2024Privacy-preserving blockchain-based federated learning for brain tumor segmentationComputers in Biology and MedicineJOURNAL↗ 2022Blockchain and homomorphic encryption based privacy-preserving model aggregation for medical imagesComputerized Medical Imaging and GraphicsJOURNAL↗ 2021Blockchain-Federated-Learning and Deep Learning Models for COVID-19 Detection Using CT ImagingIEEE Sensors JournalJOURNAL↗ 2021An Integration of blockchain and AI for secure data sharing and detection of CT images for the hospitalsComputerized Medical Imaging and GraphicsJOURNAL↗ 2019A Multimodal Malware Detection Technique for Android IoT Devices Using Various FeaturesIEEE AccessJOURNAL↗ 2019Research on Data Mining of Permission-Induced Risk for Android IoT DevicesApplied SciencesJOURNAL↗

Selected bibliography assembled from previously supplied CV data plus a repository-linked preprint. Check author attribution and DOI metadata before presenting this as a complete official publication record. Submitted work is not labeled as published.

07 / WORK TOGETHER

Research is better
when it is shared.

Interested in trustworthy agentic AI, medical intelligence, reproducibility, or collaborative research systems? I currently offer online collaboration only. See what to include in your email.

Email about collaboration ↗Connect on LinkedIn ↗Explore GitHub ↗

Intelligent systems.
Real-world problems.
Evidence that matters.

Not just a publication list.
A window into how the work operates.

Selected work.