VD
I am

Vinay Damarasing

Vinay Damarasing
DSA, distributed systems, and graph neural nets — 3rd-year CS, MLRIT. Building HTGNN-based failure diagnosis and shipping backend systems on the side. Learn more.
GitHub Activity
commits · last 12 months
Skill Matrix
Core
Applied
Stack
DSA (Graphs/DP)88
Java / Spring Boot85
Python / FastAPI87
SQL / DBMS83
System Design72
Linux / DevOps80
GNN / PyG78
RAG / Agentic AI84
REST API Design86
Docker / Podman79
AWS68
Web3 / Solidity65
IntelliJ IDEAPodmanGit / GitHub CLIPostmanMavenFedora KDESDKMAN
Featured Project

HTGNN — Cascading
Microservice Failure
Diagnosis

Writeups
CICILIYA DEBUG LOG

Fixing Fake-Tool Narration

Entry 03 — Agentic Loop
My Experience

Infosys Springboard — SentinelNet

ML-based NIDS, 92.18% accuracy · Internship

CIE MLRIT — Technical Operations Lead

Current

Major Project — HTGNN Diagnosis

Student Lead · 2026–27

Projects

01. HTGNN Failure Diagnosis

Heterogeneous temporal GNN for diagnosing cascading failures across microservices, mapping service dependency graphs over time to localize root causes faster than log-based heuristics.

TypeFinal Year Major Project
RoleStudent Lead
ToolPyTorch Geometric, Python
SkillGNN, Time-series, Graphs
ConceptRoot Cause Localization
StatusReview-0 Complete
Core Features
  • Heterogeneous service-dependency graph modeling
  • Temporal attention over cascading failure windows
  • Root-cause node ranking
  • Faculty-mentored, 3-member team
[ architecture diagram / results screenshot ]
02. Ciciliya — Agentic RAG Chatbot

Migrated from a fixed RAG pipeline to a native tool-calling agentic loop, then debugged five distinct bug classes: fake-tool-narration, entity conflation, hardcoded follow-up bypass, duplicated fallback logic, and quality-blind duplicate detection.

TypePersonal Project
RoleSolo Developer
ToolFastAPI, ChromaDB, Groq, Tavily
SkillAgentic Loops, RAG, Debugging
ConceptTool-calling Architecture
StatusNot deployed (RAM-limited)
Core Features
  • Split system prompt + dynamic context injection
  • Context-windowed entity extraction
  • Consolidated rate-limit/fallback helpers
  • Retry-aware duplicate tool-call detection
[ agentic loop diagram / repo screenshot ]
03. SentinelNet — ML-based NIDS

Network intrusion detection system built during the Infosys Springboard internship, trained and evaluated on the NSL-KDD dataset.

TypeInternship Project
RoleContributor
ToolPython, scikit-learn
SkillML Classification, Feature Eng.
ConceptIntrusion Detection
Result92.18% Accuracy
Core Features
  • NSL-KDD dataset preprocessing
  • Multi-class attack classification
  • Feature importance analysis
  • Springboard-mentored delivery
[ confusion matrix / dashboard screenshot ]
More projects (LinkForge, Cryptrack, SkyVault) can be added as 04, 05, 06 sections.