Skip to content
View Kaizer-1's full-sized avatar
9 followers · 12 following

Highlights

  • Pro

Block or report Kaizer-1

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.

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
Kaizer-1/README.md

👋 About Me

name:      Kaizer Dewaswala
role:      AI / GenAI Engineer · Backend SDE
education: B.E. Information Science & Engineering @ RVCE, Bengaluru (CGPA 8.43)
focus:     LLM orchestration · Multi-agent systems · RAG pipelines · Real-time backends
building:  Production-grade AI systems with an obsession for latency, cost & measurable outcomes
open_to:   AI/GenAI Engineer & Backend SDE roles (2026)
  • 🤖 I build agentic AI systems — multi-agent orchestration, RAG pipelines, and real-time data platforms that actually ship.
  • 📉 I care about the numbers that matter: latency, cost-per-query, and accuracy — not just "it works."
  • ⚙️ Comfortable across the stack: LLM tooling → FastAPI backends → React frontends → embedded ML.
  • ⚡ I move fast with agentic coding workflows (Claude Code, phase-structured builds).

🛠️ Tech Stack

Languages

Python C++ JavaScript SQL

AI / ML

LangChain LangGraph scikit-learn NumPy Pandas Hugging Face MCP

Backend & Frameworks

FastAPI Node.js Express React Next.js

Databases & Tools

PostgreSQL Neo4j SQLite Git Docker


🚀 Featured Projects

🧠 Company Brain

Self-updating knowledge graph that ingests company docs & Slack into Neo4j and answers multi-hop questions via a 10-route LangGraph agent.

Python · LangGraph · Neo4j · FastAPI

⚡ 9× faster ingestion (52s → 6s) · 100% routing accuracy

⚖️ LLM Council

Research study on when multi-agent LLM debate is worth its cost, with an XGBoost router that decides when to escalate.

Python · XGBoost · LangGraph · FastAPI

⚡ 91.8% accuracy at $0.0016/query across 5 benchmarks

🚨 Respondr

End-to-end emergency response platform that turns live distress calls into structured, prioritized incidents in real time.

TypeScript · Twilio · Whisper ASR · NLP

⚡ Live call → structured incident in < 5s

❤️ HRV Stress Wearable

Real-time stress-detection wearable on ESP32 with a multi-sensor HRV array (ECG/PPG/GSR/temp/IMU) + FastAPI inference.

Python · ESP32 · SVM · FastAPI · React

⚡ 0.891 LOSO accuracy across 16 subjects

More: 🔍 Code-Search — hybrid BM25 + DistilRoBERTa search (0.78 MRR)  ·  🔐 AI-Secure-Comm — TPM cryptography + AES-256-GCM + Random Forest IDS


📊 GitHub Stats



🐍 Contribution Snake

github-snake

📫 Let's build something.

LinkedIn

Pinned Loading

  1. Company-Brain Company-Brain Public

    Self-updating knowledge graph that ingests company docs and Slack into Neo4j, answers multi-hop questions via a 10-route LangGraph agent.

    Python

  2. LLM-Council LLM-Council Public

    A research study on when multi-agent LLM debate is worth its cost. XGBoost router achieves 91.8% accuracy at $0.0016/query across 5 benchmarks.

    Python

  3. Respondr Respondr Public

    Respondr is an end-to-end emergency response system that converts live distress calls into structured incidents, prioritizes them in real time, and enables operators to dispatch and track emergenci…

    TypeScript

  4. hrv-stress-wearable hrv-stress-wearable Public

    Real-time stress detection wearable using ESP32 + multi-sensor HRV (ECG / PPG / GSR / temp / IMU), SVM trained on WESAD, FastAPI inference service, and React dashboard. LOSO accuracy 0.891 across 1…

    Python

  5. Code-Search Code-Search Public

    Hybrid semantic code search engine combining BM25, neural embeddings (DistilRoBERTa), and Reciprocal Rank Fusion. Built on CodeSearchNet with 0.78 MRR.

    JavaScript

  6. ai-secure-comm ai-secure-comm Public

    Secure communication system combining Tree Parity Machine neural cryptography, AES-256-GCM authenticated encryption, and a Random Forest IDS that detects MITM and replay attacks in real time.

    Python