About Skills Infrastructure Systems Work Activity Education Hire Me
Python · AI · ML Engineer Available for Opportunities

ARPIT Bhojani ENGINEER

Building systems that think, see, and scale — fusing backend precision with production AI. Real-time surveillance meshes that identify faces in under 200 ms. Multimodal RAG pipelines digesting 1.5 lakh PDF pages at 16 pages/sec. Agentic LLM stacks that run entirely on-device, with zero API cost. Python · Django · Computer Vision · Gen AI — every day, every commit.

Scroll to explore
Arpit Bhojani
30+
Repositories
10+
Projects
8.22
CGPA
365+
Commits
Python3 Django REST Computer Vision RAG Systems LangGraph YOLOv8 Gen AI LLM Orchestration PyTorch Ollama Multiprocessing Vector DB Python3 Django REST Computer Vision RAG Systems LangGraph YOLOv8 Gen AI LLM Orchestration PyTorch Ollama Multiprocessing Vector DB
About Me

I'm the kind of engineer who can't sleep when a system is broken — and can't stop building when an idea clicks.

It started with a simple question: what if software could understand the world the way we do? That pulled me from basic Python scripts into training real computer vision models, building secure REST backends, and finally — systems that think, reason, and act autonomously.

My day-to-day is backend engineering in Django and Python — REST APIs, JWT auth, RBAC, N+1-free ORM queries, SSE streaming, and multiprocessing pipelines. On top of that I build the AI layer: Hybrid RAG pipelines (dense + sparse + RRF), agentic LangGraph graphs, multimodal document processors, and local LLM stacks via Ollama — all production-grade, not just demos.

In computer vision I work with YOLOv8, InsightFace ONNX, YOLO-World, Qwen-VL, and DeepFace — building real-time surveillance systems that process 10+ concurrent camera feeds, detect faces in under 200 ms, and stream alerts over SSE with MJPEG-compressed video at 60% bandwidth reduction.

I maintain a detailed public work log at AI-ML-at-Cactus-Creatives — daily notes, experiment results, architecture decisions, and every lesson learned the hard way. I commit every single day, not to fill a streak, but because I genuinely work every day.

System Architecture E2E Encryption CV Pipelines RAG Systems AI Agents Multiprocessing Django REST Gen AI LLM Orchestration ONNX · TensorRT Vector DB RTSP Streams SSE Streaming Playwright
30+
Repositories
10+
Projects
8.22
CGPA
365+
Daily Commits
Technical Skills

Technical Arsenal

Languages & Core
Python 3 JavaScript HTML5 CSS3 Bash SQL
Proficiency
Backend & APIs
Django DRF FastAPI Flask REST APIs JWT · BCrypt RBAC SSE AsyncIO WebSockets Celery Pydantic
Proficiency
Databases & Storage
PostgreSQL MySQL SQLite MongoDB Redis ChromaDB Qdrant SQLAlchemy
Proficiency
Computer Vision & ML
PyTorch OpenCV YOLOv8 YOLO-World InsightFace DeepFace MediaPipe ONNX TensorRT NumPy Pandas Ultralytics
Proficiency
AI · RAG · Gen AI
LangChain LangGraph Ollama HuggingFace Pydantic-AI BGE-M3 BM25 · RRF Qwen-VL Llama 3 Gemini API Claude API RAGAS Anomaly Det.
Proficiency
DevOps · Tools · Streaming
Git · GitHub GitLab BitBucket Docker Kubernetes Postman Streamlit Playwright Multiprocessing FFmpeg RTSP · MJPEG WebRTC ONVIF PyPDF · PyMuPDF UV Claude Code
Proficiency
Core Systems

Infrastructure & Architecture

01
Core Infrastructure & Hosting
  • Cloud Platforms — AWS / GCP / Azure
  • Containerization — Docker, Docker Compose
  • Orchestration — Kubernetes (basics)
  • CI/CD — GitHub Actions, Jenkins
  • Web Servers — Nginx + Gunicorn
  • IaC — Terraform, Ansible
02
Scaling & Performance
  • Load Balancing — NGINX, Round Robin, Least Connections
  • Scaling — Horizontal & Vertical
  • Database Optimization — Indexing, Sharding, Replication
  • Caching — Redis, Memcached
  • CDN — Cloudflare / AWS CloudFront
03
Advanced Backend Patterns
  • Microservices Architecture
  • Task Queues — Celery + RabbitMQ / Kafka
  • API Gateway & WebSockets
  • Rate Limiting & Throttling
  • Event-Driven Systems
04
AI & Data Systems
  • Model Serving — TorchServe, Ollama, vLLM
  • Data Pipelines — ETL, Apache Airflow
  • Object Storage — S3 / MinIO
  • Vector Databases — Qdrant, ChromaDB
05
Reliability & Security
  • Observability — Prometheus + Grafana
  • RBAC & Authentication
  • Circuit Breakers & Retry Logic
  • SSL/TLS, Encryption
  • Logging & Monitoring
06
Automation & DevOps
  • System Scripting — Python / Bash / PowerShell
  • OS Management — Linux (Debian, Ubuntu, Arch)
  • Automation — Cron, Systemd, Airflow
  • Hardware Ops — RTSP, ONVIF, Edge Gateway
  • Disaster Recovery — Backup & State Management
Architecture & Systems

How My Systems Think

INGESTION PROCESSING RETRIEVAL CORE GENERATION OUTPUT DOCUMENTS PDF PagesText + Images User QueryNatural Language Camera FeedRTSP Stream PROCESSING YOLOv8 Visual DetectionCharts · Tables · Figures Qwen-VL CaptioningAuto-caption visual regions Chunking · Tokenization RETRIEVAL CORE BGE-M3 Dense512-dim embeddingsChromaDB / Qdrant BM25 SparseKeyword matchingTF-IDF scoring RRF Fusion · Re-rankingReciprocal Rank Fusion — merges dense + sparse RAGAS EvaluationFaithfulness · Context Recall · Answer Relevancy GENERATION Llama 3 via OllamaLocal · Zero API cost LangGraph AgentStateful agentic graph 16.1 pages/sec · 8 workers OUTPUT ✓ Grounded AI Answer ✓ Source Citations ✓ Visual Context ✓ Eval Metrics
INPUT MOTION GATE NEURAL CORES IDENTITY STORE ALERT OUTPUT RTSP Camera MJPEG Stream USB / IP Cam 10+ concurrent feeds MOTION GATE Grayscale Delta160x120px · fast check VRAM Saver · 70%Reduces idle GPU load Threshold Pass/SkipGate on motion delta FACE CORE InsightFace ONNX · 512D YOLOv8 Face Detector CONTEXT CORE YOLO-World · 80+ classes ROI Tracking · CUDA IDENTITY STORE Qdrant Vector Search512D cosine similarity SQLite FallbackPersistent · survives reboot Watchlist MatchWANTED → trigger alert RBAC: Admin · Field · Worker ALERT SYSTEM SSE Dashboard · <200ms MJPG Stream · 60% BW JWT · BCrypt · RBAC ✓ Face Match · Alert Sent
CLIENTHTTP · WS · SSE AUTH LAYER JWT Tokens RBAC Roles BCrypt Hash 3-tier permissions DJANGO CORE Django REST Framework ViewSets · Serializers select_related · prefetch_related AsyncIO · SSE Streaming Multiprocessing.Pool PostgreSQLPrimary data store Vector DBChromaDB · Qdrant Cache LayerSession · Query cache AI LAYER LangGraph Agent Ollama LLM Local BGE-M3 Embedder
User QueryNatural Language LANGGRAPH STATE Intent RouterClassify: search / wiki / direct / tools WikipediaAPI lookup Web SearchDuckDuckGo CalculatorMath tools Llama 3.1 via OllamaNative tool-calling · local inference Persistent Chat Memory · State Graph REFLECTION Answer Sufficient? Re-tool if Needed Finalize Response RESPONSE ✓ Grounded Answer ✓ Tool-backed Facts ✓ Zero API cost
Technical Design Docs
RFC · Architecture Decisions · Trade-offs
Before implementing complex features, I write comprehensive design documents that outline the problem, proposed solutions, and critical trade-offs.
  • Problem Statement: Clear articulation of what needs solving and why
  • Proposed Solutions: Multiple approaches with pros/cons analysis
  • Trade-off Analysis: Why I chose NoSQL over SQL, sync vs async, etc.
  • Security Considerations: Authentication, authorization, data protection
  • Scalability Path: How the system grows from 100 to 100k users
Performance Optimizations
Benchmarks · Metrics · Before/After Analysis
Data-driven optimization with measurable results. Every improvement is benchmarked and documented.
API Response Time
-60%
Database Queries
-92%
Memory Usage
-70%
  • Optimized O(n²) logic to O(n log n) with proper data structures
  • Database indexing strategy reduced query time from 2.3s to 180ms
  • Implemented Redis caching layer for frequently accessed data
  • Lazy loading and pagination for large datasets
Infrastructure as Code
Docker · CI/CD · Cloud Deployment
Production-ready deployment pipelines with automated testing, containerization, and monitoring.
  • Containerization: Docker multi-stage builds, optimized image sizes
  • CI/CD Pipelines: GitHub Actions for automated testing & deployment
  • Cloud Infrastructure: AWS/GCP deployment with auto-scaling
  • Monitoring: Prometheus metrics, error tracking, performance logs
  • Database Migrations: Zero-downtime schema updates
API Documentation
OpenAPI · Swagger · Developer Experience
Well-documented APIs with clear schemas, error codes, and example requests make collaboration seamless.
  • OpenAPI Specs: Auto-generated Swagger UI for interactive docs
  • Clear Error Codes: Standardized HTTP status codes with detailed messages
  • Request/Response Examples: Real-world use cases in documentation
  • Authentication Guide: JWT implementation with refresh token flow
  • Rate Limiting: Documented limits and best practices
Selected Work

Projects I’ve Built

Explore GitHub

MatSetu / AI Content Mesh

Advanced AI-driven educational resource aggregator and personalized learning path generator.

Python FastAPI PostgreSQL

Sentrix AI / Surveillance Mesh

Professional distributed AI surveillance with edge face detection and centralized neural recognition.

FastAPI YOLOv8 Qdrant

Ghost Fetch / Web Automation

Silent, headless web automation tool with smart retry logic and resume support.

Python Playwright Chromium

Vision Scribe / Multimodal RAG

Production-grade multimodal RAG pipeline treating every document page as both text and image.

Django PyTorch ChromaDB

InsightHub / Blog System

Robust Django CMS for dynamic content lifecycle management, featuring threaded comments and moderation.

Python Django PostgreSQL

Nexus AI / Agentic Stack

Distributed edge-intelligence system using Flask, Pydantic-AI, and local Ollama agents.

Flask Pydantic-AI Ollama

CareerHub / Recruitment Ecosystem

Scalable recruitment ecosystem with three-tier RBAC and optimized database query patterns.

Django RBAC PostgreSQL

CryptGuard / Encrypted Messaging

E2E encrypted messaging system using proprietary 3-layer bitwise XOR encryption.

Django E2EE XOR

CampusNexus / ERP System

Comprehensive educational resource planning system for campus management.

Python Django Bootstrap

AI-Agent / Agentic Researcher

Fully local private AI agent routing between Wikipedia and live web search with persistent memory.

LangGraph Ollama Llama 3.1

CamVision / ROI Tracking

Real-time object tracking system with region-of-interest gating for optimized processing.

OpenCV YOLOv8 CUDA

AeroPiano / Hand Tracking

Interactive virtual piano controlled by real-time hand gesture recognition and skeletal tracking.

MediaPipe Python PyGame

PalmID / Biometric Auth

Secure biometric authentication system based on palm-vein and texture analysis.

Computer Vision Python PyTorch

Vision Forge / Model Studio

YOLOv8 model training studio with live epoch log streaming via Server-Sent Events.

Python Streamlit YOLOv8

MediScan / Health Analysis

AI-powered medical report analysis and diagnostic assistance tool.

Gen AI FastAPI Python
Daily Work

Contributions & Activity

Daily GitHub Commits
Consistent Daily Updates

GitHub is a living journal. Every day brings new commits — features, refactors, experiments. 30+ repositories across ML pipelines, backend systems, and AI tooling.

  • 30+ active repositories across Python, Django, ML
  • Daily commit discipline — no zero days
  • Public projects and private work tracked consistently
AI-ML at Cactus Creatives
Professional Work Log

A dedicated repo tracking real-world AI/ML work. The notes/ folder documents daily progress, models trained, and architectural decisions.

  • Daily notes in structured markdown format
  • Experiments, results and lessons documented
  • Architecture decisions captured per sprint
Learning Roadmap
Current Sprint
Computer Vision — YOLO · OpenCV · ROI Done
RAG & Agents — Pipelines · Agentic AI Done
Transformers — HuggingFace · Fine-tuning Active
LLM Orchestration — LangChain · LangGraph Active
Gen AI — Prompt Eng · Multimodal · APIs Active
MLOps — Tracking · Deployment · CI/CD Next
Deep Learning — CNNs · Attention Next
Open Source
GitHub · Fork · PR · Collaborate

Building open-source tools for the community — Python, Django, and ML ecosystem.

  • Fork on GitHub — get your own copy instantly
  • Clone locally, branch, add your fix or feature
  • Open a PR — I review all pull requests personally
  • Get merged — your contribution goes live
View GitHub
Education

Academic Background

M.Sc. in IT & CA
Saurashtra University, Halvad

Pursuing advanced specialization in IT and Computer Applications — deepening expertise in system architecture, algorithms, and applied AI/ML methodologies.

2024 — 2026CGPA 8.22
Bachelor of Computer Applications
Ganpat University, Mehsana

Foundation in computer science — OOP, DBMS, web technologies, data structures, and software engineering principles that underpin all my backend work today.

2021 — 2024CGPA 7.35
Python Programming Certification
Pransh Career and Study Point · Gandhinagar, Gujarat
Grade A+
Available for Opportunities

Let's Build
Something.

Open to backend engineering, ML engineering, and AI-native development roles. Systems thinking, security-first design, and daily-commit work ethic.

Send a Message