3D contributions · Project skyline · Research universe · Evidence · Metrics
Agentic AI engineer · Applied machine learning researcher · Robotics and navigation · Clinical AI. I build source-grounded systems that preserve state, expose uncertainty, and survive evaluation beyond polished demonstrations.
The isometric calendar is generated from GitHub's own daily contribution data. If the API cannot supply a complete period, the visualization explicitly reports the gap; it never uses fabricated activity. Inspect the underlying data.
Data source: public GitHub repository snapshot. Tower heights have a linear scale; labels show exact star counts. Private research is excluded.
Explore the research programmes and verified repositories
| Direction | Research question | Public reference |
|---|---|---|
| APEX-G / adaptive learning | How can agents generalize, persist knowledge, and improve over tasks? | Research in progress; source repository is private |
| Evidence-grounded clinical AI | How can multi-turn reasoning preserve state, provenance and safety? | Private implementation; no public repo link |
| Physics-informed estimation | Can learned state estimation satisfy kinematic constraints? | TE-PINN |
| Inertial navigation | How should learned systems estimate orientation under sensor uncertainty? | 6DoF IMU attitude estimation |
| Robotic mapping | How can systems integrate perception, localization, and mapping? | OpenSLAM, Deep-VO |
| Research operations | How can automated research workflows retain traceability? | PhD Tracker |
Research pipeline: from question to reproducible system
File markers indicate only whether documentation, tests, workflows, licence, citation, or environment files exist; they do not demonstrate empirical quality or test success.
The following panels are generated with Lowlighter Metrics when upstream rendering succeeds. Source-backed public data is displayed as a fallback if it does not.
 |  |
 |  |
Provenance, verified activity and full repository radar
- End-to-End Deep Learning Framework for Real-Time Inertial Attitude Estimation using 6DoF IMU — Measurement (2023), research code.
- Recent Advancements in Deep Learning Applications and Methods for Autonomous Navigation — survey materials.
- TE-PINN orientation estimation — experimental implementation, with no invented publication status.
Evidence before interpretation. Baselines before novelty. Explicit state before elaborate orchestration.