ANM / About
About
- AI / Machine Learning Engineer
- Bangalore, India
I'm a final-year computer science student at RV University in Bangalore, working in machine learning: computer vision, applied language research, and systems that run on local hardware.
I learn by building. Most of what is here began as a question I could only answer by making the thing and watching where it broke: a server that turns a phone's video into a 3D model you can measure, a legal study in which our own pipeline had manufactured its striking result, a guardrail that sits between me and the coding agents I use, a coach that corrects your form out loud.
This site states only what my files or public pages back up. Where a number was not measured properly, it is labeled or left out, and every project says how far it actually got. A small robot called Ember will show you around.
ANM / Method
How I work
I learn by building. Working output first, theory after.
The record, counted from this site's own data
- Projects in the line-up
- 22
- Papers
- 2
- Roles
- 2
- Awards
- 2
- Course certificates
- 8
- Workshop certificates
- 3
- Certificates of taking part
- 16
- Skills listed
- 50
- Skills tied to a project that uses them
- 37
ANM / How this site works
Mechanics
- 7 plain facts
- nothing generated, nothing sent
01
Ember, the companion
Ember is a small character drawn in your browser. Every line it speaks was written in advance and recorded as an audio file. Nothing is generated while you are here: there is no language model and no service behind it.
02
Questions
When you ask Ember a question, your words are compared, inside your browser, with a fixed set of prepared questions, and the closest prepared answer is played. The question is not sent anywhere. If nothing fits, Ember says so and points to the contact details.
03
Nothing leaves your device
The site loads no scripts, fonts, images or models from other servers. It has no analytics and no trackers, and it sets no cookies. What you type and what you choose stay in this browser. A few things are remembered in this browser's own storage and nowhere else: that you have seen the opening, the answer you gave to its first question, whether sound is off, whether Ember is asleep, the night palette, whether you have taken the tour, how many visits you have made, the last page you were on, and your progress on the project map (the places you have visited, the questions you have answered, and whether you last looked at the map or the list). For the length of a visit it also notes which remarks Ember has already made, so that it does not repeat them.
04
The voice
The voice is synthetic. It was generated ahead of time, offline, with an open speech model (Kokoro), and is served from this site as ordinary audio files. Every spoken line is also shown as text.
05
The demos and exhibits
The pose demo on the ARI Sync page starts with a simulated skeleton: nothing is downloaded and no permission is asked. If you press "Use my camera", the pose model (about 24 MB, served from this site) runs on your device, inside this tab. The camera picture is never uploaded and never stored, the pose library's own usage reporting is switched off, and the timings shown are measured on your device. The chart on the LawVerse page is drawn in your browser from the study's own result files. The project map is a drawing made when the site is built, and every place on it is an ordinary link.
06
Proof markers
Numbers, awards, certificates and publications carry a small marker. Verified: backed by a file or public page that has been checked. Owner-stated: stated by the owner; nothing contradicts it, but proof is not attached yet. Pending: stated by the owner; proof is expected but is not public or not supplied yet.
07
Motion and sound
Sound starts only after you tap, click or press a key, and can be muted at any time. If your device asks for reduced motion, the opening becomes a short fade and page content appears without movement. The résumé page has no animation and no script.