// projects — 09 builds

Systems I've Built

Production-grade AI systems and automation infrastructure, from real-time voice agents to self-hosted workflow engines.

0x01AI Voice Agent

Multilingual AI Calling Agent

With Efat Anan Shekh

Production-grade multilingual AI voice agent with native Bangla support. Handles live telephony calls with streaming speech processing, multi-turn dialogue, and real-time language switching.

// problem

Businesses in multilingual markets need voice agents that understand local languages. English-only solutions fail, and building native language AI requires specialized speech models.

// key features

Native Bangla speech recognition & synthesisMultilingual conversation supportReal-time language switchingNatural dialogue management+2 more

// stack

PythonAsteriskWebSocketLLMMultilingual TTS/STT

// outcomes

  • First production Bangla voice agent
  • Multilingual customer experience
  • 24/7 automated call handling
0x02Workflow Automation

Social Media Automation Platform

With Efat Anan Shekh

Enterprise-grade automation system for social media management, AI-powered customer engagement, lead generation, and multi-channel notification orchestration.

// problem

Manual social media operations create bottlenecks, missed leads, and inconsistent customer engagement. Teams spend hours on repetitive tasks instead of strategy.

// key features

Facebook comments auto-reply with AIMessenger AI assistant integrationLead detection & qualificationTelegram/WhatsApp instant notifications+3 more

// stack

n8nNode.jsPostgreSQLREST APIsWebhooksLLM

// outcomes

  • 80% reduction in manual tasks
  • Zero-delay lead response
  • Unified multi-platform automation
0x03Speech AI

BanglaSTT with BanglaSER

Real-time Bangla speech recognition system with speaker emotion recognition, optimized for telephony environments achieving 90%+ accuracy on live calls.

// problem

Bangla speech recognition lacks telephony-grade accuracy and emotion detection. Existing solutions fail on real-world call audio with noise and compression.

// key features

90%+ real-time telephony accuracyBangla speech-to-textSpeaker emotion recognitionTelephony noise handling+2 more

// stack

PythonPyTorchWhisperSER ModelsWebSocket

// outcomes

  • Production telephony deployment
  • Bangla language AI advancement
  • Real-time emotion insights
0x04AI Audio

Real-Time Text-to-Speech Reader

Low-latency read-aloud experience that converts selected webpage text into natural, expressive speech with seamless playback and synchronized word highlighting.

// problem

Traditional TTS systems have high latency, unnatural voices, and no visual feedback. Users need real-time, engaging audio experiences.

// key features

Real-time streaming playbackNatural expressive voice outputLive word highlighting syncSeamless continuous playback+2 more

// stack

JavaScriptWeb Audio APITTSDOM Manipulation

// outcomes

  • Instant audio feedback
  • Enhanced reading experience
  • Accessibility improvement
0x05AI Audio Platform

Twilight Story Teller

End-to-end AI storytelling platform that transforms written stories into expressive, high-quality spoken narration with custom voice creation and style control.

// problem

Creating audiobook-quality narration requires expensive voice actors and studios. Content creators need accessible, customizable voice generation.

// key features

Story-to-voice generationExpressive delivery control (calm, dramatic, playful)Custom voice creation pathwayDesktop GUI for rapid iteration+2 more

// stack

PythonPyTorchTTS ModelsGradioCLI Tools

// outcomes

  • Professional narration quality
  • Custom voice branding
  • Scalable content production
0x06Desktop AI

Hybrid Multilingual Transcription Studio

GPU-accelerated desktop application for private, offline transcription of meetings and desktop audio with speaker-aware segmentation and multilingual support.

// problem

Cloud transcription tools create privacy, compliance, and reliability risks. Organizations need local-first alternatives that handle noisy, mixed-language conversations.

// key features

Mic, browser tab & app audio captureSpeaker-aware transcript structuringMultilingual & mixed-language supportNoise & low-volume speech enhancement+2 more

// stack

PythonPyTorchCUDAPyQt6WhisperNeMoFFmpeg

// outcomes

  • Zero cloud dependency
  • Reliable meeting transcription
  • Production-ready desktop build
0x07Desktop AI

Offline Speech Processing System

GPU-accelerated desktop application for offline speech recognition with speaker diarization and neural noise suppression.

// problem

Cloud-based speech processing raises privacy concerns and requires constant connectivity. Organizations need offline solutions.

// key features

High-accuracy offline recognitionMulti-speaker identificationNoise suppressionWord-level timestamps+2 more

// stack

PythonPyTorchCUDAWhisperPyInstaller

// outcomes

  • Complete data privacy
  • No cloud dependency
  • Production-ready installer
0x08SaaS Platform

RosterBhai - Team Roster Management

With Efat Anan Shekh

Complete roster management system with real-time updates, shift swapping, and team scheduling. Built for efficiency and mobile accessibility.

// problem

Manual scheduling is time-consuming, error-prone, and creates communication gaps. Teams need centralized, real-time roster management.

// key features

Drag-and-drop schedule builderOne-click shift swap requestsReal-time team notificationsMobile-friendly staff access+2 more

// stack

Next.jsTypeScriptPostgreSQLReal-time Sync

// outcomes

  • 70% faster scheduling
  • Reduced scheduling conflicts
  • Improved team coordination
0x09Healthcare SaaS

DocSaaS - Smart ePrescription

With Efat Anan Shekh

Modern ePrescription platform with built-in medical database, patient analytics dashboard, and practice insights for healthcare providers.

// problem

Paper prescriptions are error-prone and don't provide practice insights. Doctors need digital tools with medical intelligence.

// key features

Smart prescription builderBuilt-in medicine databasePatient analytics dashboardPractice trend insights+2 more

// stack

Next.jsTypeScriptPostgreSQLAnalytics

// outcomes

  • Faster prescriptions
  • Data-driven practice insights
  • Improved patient tracking
~/deploy — run

$ ./start_project.sh

Ready to Automate?

Let's build intelligent systems that transform your operations. From AI voice agents to workflow automation, I deliver production-ready solutions.

Start a Project
Production-Ready SystemsEnd-to-End ImplementationScalable Architecture