kaushalrog (Kaushal Reddy) · GitHub
Skip to content
View kaushalrog's full-sized avatar
🎯
Focusing
🎯
Focusing
36 followers · 34 following

Highlights

  • Pro

Block or report kaushalrog

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
kaushalrog/README.md
Blueprint drawing KR-001: Kaushal Reddy, one engineer across three domains (AI systems, security, edge computing). A schematic shows a model that acts when sure and abstains when not. Note 1: a model must know when it is wrong.

PORTFOLIO kaushalrog.github.io/portfolio  ·  LINKEDIN kaushal-s  ·  PAPER IEEE Xplore

Note

The demo always works. I build for the run after the demo. I'm Kaushal. I work where AI, security and edge hardware meet: models that admit when they're unsure, intrusion detection that watches how a system behaves instead of what it's sent, and code small enough to run on a Raspberry Pi.

SHEET 2  Open questions

Featured project KR-100, Satellite Optical Fault Recovery: when a laser link between satellites fails, what broke, and can it be fixed? MATLAB and Simulink, 4 to 10 satellites, 320 matched trials. Detection accuracy 72.1% to 96.5%, exact fault isolation 9.5% to 71.5%, false-alarm rate 33.4% to 3.0%. Reproduced 320 of 320 trials exactly.

Latest: ▶ Live simulator  ·  MATLAB results  ·  Repository
Also live: Spacecraft formation attitude-control digital twin

WDS-IDS, research: can you catch an attack without reading the payload? Method: OS behaviour and drift. Data: SRBH-20, 5-fold CV. Latency: under 12 ms on a Raspberry Pi 5. SciRAG-UQ, building: when should a model say I don't know? Method: multi-signal confidence. Output: answer or abstain. Target: BDA 2026.
TrafficFlow AI, accepted: what if a traffic light could read the queue? Method: machine learning plus fuzzy logic. Live congestion dashboard. Accepted at ICPEEV 2026. MCDIP-ADMM, research: how much of a CT scan survives less radiation? Method: ADMM plus multi-code Deep Image Prior. Evaluated with PSNR and SSIM on synthetic data. Unsupervised.

Select a sheet to open its repository. Off the drawing board: AntBot, a hexapod surveillance robot, published on IEEE Xplore.

SHEET 3–4  Live telemetry

Automatically updated counts of public repositories, stars on my repositories, and followers. Language distribution by code bytes across my public, non-fork, non-archived repositories. Top six shown; this measures code volume, not proficiency.
Change log · latest public activity
Date (UTC) Activity Repository
2026-10-21 Star · Public activity harry0703/MoneyPrinterTurbo
2026-10-20 Star · Public activity 666ghj/MiroFish
2026-10-20 Pr · opened #2 Rishithachappidi/Active-Fault-Diagnosis-an…
2026-10-20 Pr · opened #2 Rishithachappidi/Quantum-Circuit-Optimizat…
2026-10-19 Push · Pushed code kaushalrog/AG-DATA
Project pulse · last 30 days
Project Your commits · last 30 days Latest push (UTC)
WDS-IDS 0 2026-08-27
SciRAG-UQ 0 2026-07-02
TrafficFlow 1 2026-09-23
CT Reconstruction 1 2026-10-14

My commits on each project's default branch. A quiet repository can be finished work; activity is not a quality score.

SHEET 5  Bill of materials

Toolbox. Languages: Python, C++, Java, SQL, Bash. Machine learning: PyTorch, TensorFlow, scikit-learn, LightGBM. LLM systems: RAG, Transformers, agentic AI, MCP. Retrieval: FAISS, ChromaDB. Systems: Linux, Docker, FastAPI, REST APIs, GitHub Actions. Security: behavioural analysis, intrusion detection. Hardware: Raspberry Pi, Arduino.

General notes

  1. Off the clock I shoot campus events with Team Media at Amrita.
  2. Badminton, between builds.
  3. I read papers until one of them turns into an experiment.
  4. Building something thoughtful in AI or security? Send a note on LinkedIn →

Drawing sheets 3–4 and the change log are regenerated by GitHub Actions hourly, on profile-repo pushes, or on demand · latest snapshot: 2026-10-21T13:47:40Z · public data only.

Pinned Loading

  1. ids---System ids---System Public

    This project introduces a new approach: instead of analyzing web request content, the system detects attacks by monitoring the operating system–level behavioral changes caused by malicious activity.

    Python 2 1

  2. mcdip-admm-ct-reconstruction mcdip-admm-ct-reconstruction Public

    An unsupervised CT image reconstruction project using the ADMM optimization framework combined with Deep Image Prior (DIP). The approach employs a Multi-Code DIP ensemble to reduce overfitting and …

    Jupyter Notebook 3

  3. ANTBOT---The-Hexapod-Robot ANTBOT---The-Hexapod-Robot Public

    Forked from divya-rithanyaS/ANTBOT---The-Hexapod-Robot

    Low-cost 21-DOF hexapod robot for surveillance with inverse kinematics-based locomotion and real-world validation.

    1 1

  4. scirag-uq scirag-uq Public

    SciRAG-UQ: Confidence-Calibrated Multi-Source Retrieval-Augmented Generation for Scientific Literature Synthesis

    Python 1 1