Design, Build, AI, Automation
Project Overview
Rechitta is a cinematic, feature-film-directed digital web experience engineered for luxury real estate developers in Dubai. It serves as a unified digital briefing engine connecting property developers with over 40,000 registered brokers across 6 international financial hubs at the exact same instant: 04:11 UTC.
The landing experience advances through an interconnected five-act sequence: Dawn, the Executive Boardroom, the Broker Terrace, the Synchronised World Chronometer, and the Dusk Horizon Loop.
The Problem and Industry Context
In rapid-growth luxury property capitals like Dubai, developers launch multi-billion-dollar landmark towers. However, information distribution traditionally relies on outdated and manual workflows:
Network Congestion: Over 40,000 registered brokers operating across more than 1,200 independent agencies.
The Scheduling Bottleneck: A standard 30-person developer sales team requires 3 to 6 months of repetitive boardroom pitches, phone campaigns, and slide presentations just to brief the local brokerage network.
Language Degradation: Cross-border buyers and international brokers speaking Mandarin, Russian, Arabic, Hindi, or French rely on third-party translation relays, causing severe loss of technical nuance, misunderstandings around payment plans, and delayed transaction closings.
From an engineering perspective, most scroll-driven video websites attempt to bind scroll offsets directly to the video currentTime property. This introduces decoder starvation, dropped frames, scroll judder, and severe performance issues on mobile devices. Rechitta required an ultra-luxury, fluid 60fps interactive narrative without compromising performance.
Key Engineering Architecture and Implementation
1. Discrete Beat Engine and Hardware GPU Decoding
To eliminate scroll judder and decoder stalls:
Developed a custom input state machine (director.ts) driven by an Intent Accumulator. User scroll wheel and touch gesture events accumulate until crossing a defined threshold before committing to a camera move.
Added cooldown guards and idle reset timers to prevent momentum overshoots.
Transitions decode short video sequences at native hardware GPU speeds (1.0x to 3.6x) rather than scrub-seeking raw video elements, delivering a locked 60fps playback experience across both desktop and mobile devices.
2. Mathematical 3D Perspective Homography Warp Engine
To render live, functioning web applications inside smartphones and boardroom monitors within moving cinematic footage:
Built a custom 8-degree-of-freedom perspective homography solver using Gaussian elimination.
Tracked normalised video coordinates across corner vertices (Top-Left, Top-Right, Bottom-Right, Bottom-Left) relative to viewport cover crops.
Dynamically calculated CSS matrix3d transformations in real time, allowing interactive application interfaces and iframes to naturally track moving physical camera angles.
3. Luminous 3D Orb Navigator and WebGL Globe
Built a celestial 3D Orb using Three.js and React Three Fibre that guides the visitor through the narrative.
Anchored orb trajectories directly to video clip timestamps rather than scroll percentages, ensuring spatial synchronisation regardless of responsive layout changes.
Implemented an interactive 3D WebGL Earth featuring custom atmospheric GLSL Fresnel glow shaders, day/night texture blending, and dynamic marker coordinates for global financial hubs.
4. Synchronised World Chronometer (04:11 UTC) and In-Memory Video Streamer
Realised the core philosophy of "The Same Moment" by mathematically calculating simultaneous local times across 6 world capitals (London, Paris, Moscow, Riyadh, Mumbai, Shanghai) from a single 04:11 UTC baseline.
Engineered an in-memory RAM Blob pre-warming engine (CityBackdrop.tsx) that precaches high-bitrate drone cinematography loops to ensure instantaneous switching between hubs with native localised typography (supporting both LTR and RTL scripts).
5. Automated Multi-Tier Video Encoding Pipeline
Built an automated multi-tier encoding pipeline using FFmpeg to support broad device compatibility:
AV1 (10-bit) for modern Chromium browsers.
HEVC / Apple VideoToolbox (10-bit) for hardware-accelerated Safari and iOS WebKit rendering.
H.264 (8-bit) as a universal baseline fallback.
Discretised 60fps WebP keyframes for zero-latency mobile first-frame underlays.
6. Mobile Polish and Continuous Infinity Loop
Eliminated video end-of-clip flickering and playback hangs on iOS and Android devices through preloaded dual-video decoders and state cleanups.
Designed an architectural infinity loop where the Act V dusk horizon smoothly cross-dissolves backwards into the Act I dawn sky for an uninterrupted, continuous presentation cycle.
Design System and Visual Standards
Palette: Deep Obsidian (#0A0A09), Surface Obsidian (#141413), Desert Gold (#C5A572), Electric Brand Blue (#3D6FF5), and Warm Off-White (#FAFAF9).
Typography: Architectural serif headings paired with clean Swiss grotesque body typography and monospace UTC telemetry clocks.
Micro-Interactions: Magnetic custom cursors, glassmorphic HUD status overlays, and fluid layout progress rails.
Key Results and Deliverables
Consistent 60fps Playback: Fully resolved video judder and decoder starvation across low-power mobile devices and high-end 4K workstations.
Cross-Platform Parity: Native hardware video acceleration and responsive touch gestures across iOS, Android, macOS, and Windows.
Tested Reliability: Automated unit test coverage powered by Vitest verifying core algorithms including homography matrix calculations, input accumulators, and timeline synchronisation.
Luxury Standard: Delivered a high-performance web experience matching the calibre and architectural scale of Dubai's premier property developers.
Technology Stack
Frontend: Next.js 16 (App Router), React 19, TypeScript
Styling: Tailwind CSS v4, Vanilla CSS Modules
3D and Animation: Three.js, React Three Fibre, Drei, GSAP, Lenis Smooth Scroll
Shaders and Math: Custom GLSL Shaders, 8-DOF Homography Solver (matrix3d)
Media Pipeline: FFmpeg (AV1, HEVC, H.264, WebP)
Quality Assurance: Vitest, ESLint
3
0
142
Full-stack developer & product designer | AI workflows
Full Stack AI Engineer |AI Agents| Generative AI |Automation
Full Stack AI Engineer |AI Agents| Generative AI |Automation
Developed an AI-powered Technical Interviewer using Tavus avatars for realistic, human-like interactions. I integrated LiveKit for video room management and agent handling, while building a FastAPI backend to handle session creation, tokens, and interview orchestration. I also worked with STT (Deepgram), TTS (Cartesia), LLMs (OpenAI/Groq), and custom voice pipelines to enable seamless real-time conversations.
0
183
Developed an AI-powered Marketing Campaign Automation platform that helps businesses create, manage, and optimize marketing campaigns across multiple channels with minimal manual effort. The solution leverages Generative AI and conversational AI to streamline campaign planning, content generation, audience targeting, and performance analysis. The platform enables marketing teams to launch campaigns faster while improving engagement and conversion rates. Inspired by modern AI-driven marketing approaches that automate campaign workflows and personalized customer interactions.
5
3
454
Designed and developed a bilingual (English/Arabic) real-time AI Voice and Chat Agent for ILAF Insurance to automate motor accident claim reporting and document intake (POC: https://ilaf-frontend.vercel.app). Built a low-latency WebRTC voice pipeline using LiveKit Agents, OpenAI Realtime API, Silero VAD, and Deepgram. Developed an asynchronous FastAPI backend with automated Civil ID policy verification, garage assignment, multi-document uploads (Police reports, licenses, damage photos), and automated email notifications, paired with a modern React 19 frontend.
0
194
Solara AI is an end-to-end voicebot and web chatbot for dental practices (Live Demo: https://solara.supermia.ai/). Built using LiveKit SIP, OpenAI GPT-4o, ElevenLabs, and FastAPI, it handles inbound/outbound phone calls and website chat. It answers patient queries, checks real-time calendar availability, and books appointments on Google Calendar. Featuring noise cancellation and WebSockets, it provides sub-second voice latency and natural conversation flows to automate dental clinic scheduling 24/7.
0
196
Flutter & full-stack dev shipping AI-powered apps.
AI Developer & AI Agent Engineer
DocFlow AI – Intelligent Document Processing Platform
DocFlow AI is an enterprise document intelligence platform that transforms unstructured documents into searchable, actionable knowledge.
The platform automatically extracts information from PDFs, Word files, invoices, contracts, and reports, builds a searchable knowledge base, and enables users to interact with their documents through an AI assistant using Retrieval-Augmented Generation.
It combines OCR, structured extraction, RAG, AI agents, and workflow automation to simplify document-heavy business operations.
Features:
PDF Processing
OCR Pipeline
RAG
AI Chat with Documents
Entity Extraction
Document Classification
Knowledge Base
Citation Generation
Role-based Access
FastAPI Backend
Tech Stack:
Python
FastAPI
LangGraph
LangChain
ChromaDB
PostgreSQL
Docker
Redis
OpenAI
OCR
0
55
ResearchX – Autonomous Deep Research AI Agent
ResearchX is a production-ready autonomous research agent that performs end-to-end research with minimal human input.
Instead of simply answering questions, the agent plans research tasks, searches multiple trusted sources, evaluates information quality, retrieves supporting evidence, and generates structured reports with citations.
Built using a multi-agent architecture, the system combines planning, retrieval, tool calling, memory, and iterative reasoning to handle complex research tasks while reducing hallucinations through grounded retrieval and verification.
Features:
Autonomous Planning
Multi-Agent Workflow
Web Search Integration
PDF Analysis
RAG Pipeline
Citation Generation
Long-Term Memory
Report Generation
Source Verification
Evaluation Framework
Tech Stack:
Python • LangGraph • FastAPI • Tavily • ChromaDB • PostgreSQL • Redis • Docker • OpenAI • Claude
0
126
Scopio – Multi-Agent AI Sales Automation Platform
Scopio is a production-ready multi-agent AI platform designed to automate B2B sales and lead generation.
The platform coordinates multiple AI agents to perform end-to-end sales workflows:
Discover potential businesses from multiple sources
Research company websites and public information
Analyze business context using LLMs
Generate personalized outreach emails
Manage follow-up workflows automatically
Maintain conversation history and workflow state
I designed the agent architecture using LangGraph, enabling planning, tool calling, memory, and multi-step execution while integrating external APIs and backend services for reliable automation.
Key Features
Multi-Agent Architecture
Autonomous Business Research
AI-Powered Personalized Outreach
Tool Calling & API Integration
Workflow Orchestration
Long-Term Memory
PostgreSQL & Redis Integration
Dockerized Deployment
FastAPI Backend
Tech Stack
Python • LangGraph • LangChain • FastAPI • PostgreSQL • Redis • Docker • OpenAI • MCP • REST APIs
0
57
What happens after someone asks your business, “Can you help with my project?”
I designed this AI lead qualification concept to show how an interior design studio could turn a website conversation into an organized enquiry, ready for its team to review.
The experience brings together:
• A conversational assistant that collects project requirements.
• Structured details including location, budget, style, and timeline.
• A dashboard that gives the owner a clear enquiry summary.
• An AI-drafted response with human approval.
The goal: give potential customers a welcoming first interaction and give the business a clearer starting point for follow-up.
This animated showcase explores the product’s UI/UX and intended workflow using GSAP and HyperFrames.
Created by Kazi Eajaj Hossain ·
Concept demo with fictional data.
Have a repetitive enquiry or customer support workflow? Let’s discuss how a custom AI assistant could fit your business.
#AIAgents #AI Automation #ConversationalAI #UIDesign #WorkflowAutomation
0
31
Algo Trading Developer - Pine Script,MQL5 &Python Automation
Designed and implemented a robust webhook-based trading automation system that connects TradingView signals to real-time trade execution via the Tradovate API. The system is built using Python and handles signal reception, validation, and order execution in a seamless automated pipeline.
The architecture ensures low-latency communication, secure data handling, and reliable execution of trades without manual intervention. It processes incoming webhook alerts, applies validation logic, and triggers precise API-based trade execution.
Key features include:
• Real-time webhook signal processing (TradingView alerts)
• Python-based signal receiver and validation engine
• Secure and reliable API integration (Tradovate)
• Low-latency execution pipeline
• Fully automated trade workflow (zero manual input)
• Scalable design for multi-strategy integration
This solution is ideal for traders and firms looking to automate their strategies using webhook-driven architecture, enabling faster execution, reduced human error, and consistent performance in live trading environments.
1
239
Developed an automated options trading bot using Python that executes trades on Deriv based on real-time strategy signals and multi-condition validation logic. The system is designed to operate continuously with minimal latency, ensuring fast and accurate trade execution in volatile market conditions.
The bot integrates signal generation, validation, and execution into a single automated workflow. It evaluates multiple technical indicators, applies strict entry conditions, and triggers trades only when high-probability setups are detected.
Key features include:
• Fully automated options trading (Deriv API integration)
• Real-time signal processing and execution
• Multi-indicator validation logic for higher accuracy
• Risk management controls and trade filtering
• Continuous operation with minimal manual oversight
• Scalable architecture for strategy customization
This solution enables traders to automate their options trading strategies end-to-end, improving consistency, speed, and discipline while eliminating emotional decision-making.
1
185
Developed a fully automated trading system for MetaTrader 5 (MT5) that executes trades based on validated multi-indicator signals. The system integrates TradingView-generated signals with a Python-based processing engine, which evaluates multiple technical conditions before triggering precise trade execution.
The Expert Advisor (EA) is designed with advanced trade logic, risk management, and real-time execution capabilities. It ensures consistent performance by removing emotional decision-making and strictly following predefined rules.
Key features include:
• Automated trade execution on MT5 (buy/sell orders)
• Integration with external signal sources (TradingView via webhook)
• Multi-indicator validation (RSI, EMA, ADX, Volume, etc.)
• Built-in risk management (lot sizing, SL/TP control)
• Real-time processing with minimal latency
• Fully customizable strategy logic
This system enables traders to automate their strategies end-to-end — from signal generation to execution — ensuring speed, accuracy, and consistency in live market conditions.
1
197
Engineered a high-performance Python-based trading signal engine designed to generate precise and reliable trade signals using multi-indicator confluence logic. The system evaluates over 12 technical indicators including RSI, EMA, ADX, Volume, and trend-based conditions to identify strong market opportunities while effectively filtering out noise and false signals.
The architecture is built to process real-time market data, validate multiple conditions simultaneously, and trigger only high-probability trade setups. This significantly improves signal accuracy and reduces unnecessary trades.
Key capabilities include:
• Multi-indicator confluence engine (12+ conditions)
• Real-time signal processing and validation
• False signal reduction using layered logic
• Scalable architecture for integration with trading APIs
• Seamless compatibility with automated execution systems (webhooks, APIs)
This system can be integrated with platforms like TradingView, Tradovate, or custom trading environments to enable fully automated trading workflows — from signal generation to execution.
Designed for traders and firms looking to achieve consistent, rule-based, and emotion-free trading through intelligent automation.
2
196
Full-Stack Engineer | Web Apps & Website Fixes
Full Stack & AI Engineer | Python, React & LLMs
Autonomous AI CRM & Revenue Intelligence Platform
Built an enterprise-grade AI-powered CRM and Revenue Intelligence SaaS platform featuring autonomous prospecting, predictive deal analytics, and automated SDR workflows.
Core Architectural Modules:
• AI Prospector (Autonomous Lead Discovery): Real-time map discovery (Google Places/OpenStreetMap) and automated web enrichment with LLM fit scoring (1-10) and custom pitch generation.
• Shadow SDR: Asynchronous cold outreach email draft generator powered by FastAPI BackgroundTasks and local/cloud LLM chaining (Ollama/Groq/Gemini).
• Dynamic CRM Pipeline: Drag-and-drop Kanban board (@dnd-kit) with strict phone deduplication, server-side pagination, and one-click lead-to-sale transitions.
• Predictive Revenue Engine: Scikit-Learn regression models for deal win-probability and cash flow forecasting.
• Conversational Sales Assistant: Integrated floating AI chatbot for natural-language pipeline queries and call log summarization.
Tech Stack:
• Backend: Python (FastAPI), SQLAlchemy (Async), PostgreSQL 15, Redis, Alembic, Docker
• Frontend: React 18, TypeScript, Vite, TailwindCSS, Shadcn UI, Zustand, TanStack Query v5
• Machine Learning & AI: Scikit-Learn, LangChain, Groq, Gemini, Ollama
0
55
Just launched a turnkey AI Media Intelligence & Growth Operating System for creator agencies, growth teams, and founders.
Most creator tools are either generic ChatGPT wrappers or lock you into expensive monthly seats with API markups.
I built an end-to-end media OS that gives you full ownership:
🚀 Core Modules:
• Deep Channel Audit: Automated diagnostics on hook retention, traffic leaks, and metadata flaws on any public or connected YouTube channel.
• 30-Day Growth Roadmap: AI-powered, data-backed content cadence, video outlines, and hook frameworks.
• Multi-Channel Sync: YouTube & Instagram analytics aggregation in one workspace.
• BYOK Architecture: Bring-Your-Own-Key LLM integration (OpenAI/Anthropic/Gemini) with zero platform markup.
• Complete Workspace: Script pipeline, blog repurposing, and LinkedIn post generation.
📦 What’s included in the Turnkey Setup:
Full commercial source code license (React UI + FastAPI backend).
Hands-on cloud deployment to your server (VPS/AWS).
OAuth & API keys configuration.
1-on-1 architecture walkthrough call.
Check out the full service offering and book a 15-min live demo walkthrough here:
https://contra.com/s/L2UlZFvm-turnkey-ai-media-os-and-you-tube-audit-system-custom-setup
0
53
Built an end-to-end AI-powered Content Planner and YouTube and Instagram Analytics platform designed to automate channel audits and streamline production workflows.
Key Highlights:
• Automated Channel Auditing: Evaluates video performance, pacing, and retention drivers.
• AI Content Strategy: Generates data-backed content outlines, hooks, and topic clusters.
• Real-time Analytics Dashboard: Aggregates core YouTube performance metrics to deliver actionable publishing insights.
Tech Stack: Python, FastAPI, React, AI/LLM integrations, TailwindCSS.
0
79
Multimodal Video SEO & Content Intelligence Engine
Engineered an end-to-end, 100% offline Multimodal Video SEO & Strategy Intelligence system designed for high-volume content creators and media marketing agencies.
Core Multimodal AI Pipeline:
• Audio Intelligence (Whisper): Millisecond-precision speech-to-text, tone detection, cadence tracking, and keyword density analysis.
• Computer Vision & Thumbnail Scorer (OpenCV + Vision LLMs): Semantic scene identification, on-screen text extraction, and predictive A/B thumbnail clickability scoring.
• Strategic Logic & SEO Pack: Correlates audio and visual signals to compute weighted retention scores (Hook 35%, Keyword 25%, Visual 25%, Sentiment 15%) and automated metadata packages.
• Competitor Benchmarking: Automated transcript gap analysis to identify underserved search demand and audience retention triggers.
• Automated Client Reporting: Production-grade automated PDF audit generation with visual breakdown charts.
Tech Stack:
• Core: Python, FastAPI (Async Orchestration), OpenCV, FPDF2, SQLite
• AI & Local Models: OpenAI Whisper, Gemma / Local LLMs via Ollama, TextBlob
• Dashboard: Streamlit Analytical Interface
0
21