┌─────────────────────────────────────────────────────────────┐
│  AI/ML Engineer • Systems Architect • Backend Developer     │
│  Building high-performance systems from first principles    │
└─────────────────────────────────────────────────────────────┘

Languages

Python      ██████████████████░░  90%    C++  ████████████████░░░░  80%     Typescript   ██████████░░░░░░░░░░  40%
JavaScript  ██████████████░░░░░░  70%    Go   ██████░░░░░░░░░░░░░░  30%

Systems & Infrastructure

┌─ Backend ──────────────────────────────────────────┐
│ FastAPI • gRPC • Node.js • NGINX • WebSockets      │
├─ Orchestration ────────────────────────────────────┤
│ Docker • Kubernetes • Microservices                │
├─ Databases ────────────────────────────────────────┤
│ PostgreSQL • MongoDB • Redis • Vector DBs          │
└────────────────────────────────────────────────────┘

AI/ML Pipeline

┌─ Frameworks ───────────────────────────────────────┐
│ PyTorch • TensorFlow • vLLM Inferencing            │
├─ Specialized ──────────────────────────────────────┤
│ CUDA • Distributed Inferencing • Benchmarking      │
├─ Applications ─────────────────────────────────────┤
│ LLMs • RAG Systems • Fine-tuning                   │
└────────────────────────────────────────────────────┘

A GPU-aware LLM benchmarking orchestration platform built on top of vLLM, designed for real-world system behavior, concurrent execution, and clean architecture design.

• Remote benchmark job submission
• Thread based Task Scheduler for parallel execution
• Cluster management for available GPU cluster
• Real-time monitoring & logs
• Structured metrics extraction
• Per-user history tracking

A large-scale production-grade, high-performance reverse proxy designed to operate under configurable latency and throughput constraints. This system provides a pluggable architecture for sophisticated traffic distribution and real-time observability.

• Tech: Node.js, http API, zod, vanilla JS

Scalable 2 server-based architecture for production-based secure wiping.

• Custom token based auth between proxy and microservices.
• Remote Wiping tool functionality for Linux Systems.
• Professional Logging system and Dashboards
• Tech: Node.js, FastAPI, Nest.js, MongoDB

Built encoder-decoder mdoel Transformer and MiniGPT implementation in pure PyTorch

• Custom attention mechanisms, positional encoding, layer normalization
• Training pipeline for dialogue datasets
• Tech: PyTorch, NumPy

Production RAG systems for PDF processing and real-time chat

• Vector similarity search with FAISS
• YouTube transcript extraction and semantic search
• Tech: LangChain, ChromaDB, FastAPI

Custom machine learning library with fundamental algorithms implemented from scratch

• Linear/Logistic Regression, Decision Trees, Random Forests, SVM
• Neural Networks with backpropagation, K-Means, PCA
• Pure Python/NumPy implementations for research clarity
• Tech: Python, NumPy, Mathematical optimization