I build LLM-powered systems and full-stack web applications, from real-time AI agents to production backends, based in Karachi, Pakistan.
constdeveloper={name:'Sameer Shah',location:'Karachi, Pakistan',skills:['Python', 'FastAPI', 'Django', 'Next.js', 'LLM Integration', 'Claude API', 'Gemini API', 'RAG Systems', 'FAISS', 'Playwright', 'Docker', 'PostgreSQL', 'Supabase'],currentFocus:'AI-powered automation & LLM products',openTo:'Remote AI Engineering & Freelance',hireable:function() {returntrue;},};Summary
I'm an AI and Full-Stack Engineer based in Karachi, Pakistan, specializing in LLM-powered applications, intelligent automation pipelines, and full-stack web systems. I build end-to-end AI products — from real-time voice agents and RAG chatbots to automated content pipelines and multi-tenant SaaS platforms. My work spans Anthropic Claude, Google Gemini, FastAPI, Next.js, and production scraping infrastructure. I'm currently open to remote freelance engagements in AI integration, backend automation, and full-stack development.
(Jun 2026 - Present)
Agentic AI Engineer
Co-Ventech
Built a multi-tenant chatbot SaaS platform (Express, TypeScript, Prisma, PostgreSQL) with Meta OAuth integration for WhatsApp, Messenger, and Instagram, letting businesses manage AI chatbots across platforms from one dashboard. Built an automated reel generation pipeline, orchestrating Gemini and Pexels via Make.com to render narrated videos. Automated lead capture, follow-ups, and business workflows using Lindy and GoHighLevel (GHL).
(Feb 2026 - Apr 2026)
AI/ML Developer Intern
Tamatos
Designed AI-powered automation systems to optimize data processing and compliance workflows. Engineered automated ETL and OCR validation pipelines for accurate, large-scale data verification. Built CRM-based lead automation systems using GoHighLevel (GHL), including lead capture, classification, and automated follow-up workflows.
(2026 - Present)
AI & Full-Stack Engineer
Freelance
Delivering LLM-powered applications and full-stack web platforms. Focus areas include AI/LLM integration (Claude, Gemini APIs), RAG pipeline development, and backend/frontend systems with FastAPI, Django, and Next.js for independent clients.
Chatbot SaaS Platform
constproject={name:'Chatbot SaaS Platform',tools: ['TypeScript', 'Express', 'Prisma', 'PostgreSQL', 'Next.js', 'Meta OAuth', 'DeepSeek API],myRole:Full-Stack / Backend Engineer,Description: Built a multi-tenant REST backend enabling businesses to connect messaging platforms via Meta OAuth and manage AI chatbot services from a centralized dashboard, with all data scoped per organization. Integrated the DeepSeek API as the LLM powering automated chatbot responses. Designed a request-and-approval pipeline where connection requests trigger emailed one-time review links, allowing admins to approve, reject, or request further information before activation. Implemented JWT authentication (Bearer token and httpOnly cookie), role-based access control, and admin APIs for organization management, KPIs, and activity monitoring.,};ClassSense: Classroom Engagement & Emotion Monitoring
constproject={name:'ClassSense: Classroom Engagement & Emotion Monitoring',tools: ['Python', 'FastAPI', 'PyTorch', 'EfficientNet-B0', 'OpenCV', 'WebSockets', 'JWT],myRole:AI/Backend Engineer,Description: Built a FastAPI backend that analyzes classroom video in real time to infer student engagement, emotion, and gaze using a PyTorch EfficientNet-B0 model. Implemented a WebSocket streaming endpoint that processes live webcam/RTSP frames with background-threaded inference. Designed a privacy-first pipeline that discards raw video and stores only per-frame analytics, with automated PDF/CSV session reports.,};AI Cold-Calling Agent
constproject={name:'AI Cold-Calling Agent',tools: ['TypeScript', 'Twilio Voice API', 'Google Gemini Live API', 'PostgreSQL', 'Prisma', 'Redis', 'BullMQ],myRole:AI Infrastructure Engineer,Description: Built an outbound calling system that dials leads via Twilio, detects voicemail vs. human pickup (AMD), and bridges live calls to Google Gemini Live for real-time voice conversation. Implemented a Redis/BullMQ job scheduler for call dispatch with retry backoff, and enabled the AI agent to book meetings mid-call via Gemini tool calling.,};AI Content Synthesis Engine
constproject={name:'AI Content Synthesis Engine',tools: ['Python', 'Gemini API', 'Playwright', 'MoviePy', 'FastAPI', 'Google Drive API],myRole:AI Automation Engineer,Description: An automated end-to-end video production pipeline. Discovers viral topics by scraping Google Trends and Reddit, uses Gemini Flash to write scripts and captions, generates cinematic background images via SiliconFlow/FLUX, renders a 15-second portrait reel with MoviePy, and uploads to Google Drive — fully automated with zero human input.,};Enterprise AI Voice Agent
constproject={name:'Enterprise AI Voice Agent',tools: ['Twilio', 'Deepgram', 'Groq', 'FastAPI', 'WebSockets', 'Python],myRole:AI Infrastructure Engineer,Description: A low-latency telephony AI infrastructure for autonomous phone call handling. Twilio Media Streams captures real-time audio, Deepgram Nova-2 transcribes it, Groq Llama-3.3-70b generates responses, and Deepgram Aura converts them back to voice — all under 1 second latency. Includes emergency triage that detects critical phrases and redirects calls automatically.,};Event Feedback Intelligence System
constproject={name:'Event Feedback Intelligence System',tools: ['Next.js', 'FastAPI', 'Gemini 2.5', 'Supabase', 'Gmail SMTP', 'TypeScript],myRole:Full-Stack / AI Engineer,Description: A full-stack AI platform for event organizers. Collects attendee feedback, runs it through Gemini 2.5 for sentiment analysis and urgency scoring, auto-generates branded PDF reports, delivers them via email, and persists everything to Supabase — all triggered automatically on each form submission. Includes an analytics dashboard with date-range bulk reporting.,};Pakistan Influencer Discovery Engine
constproject={name:'Pakistan Influencer Discovery Engine',tools: ['Python', 'Playwright', 'Instagram Web API', 'TikTok SIGI', 'Flask', 'SQLite],myRole:Data Engineering & Scraping,Description: A high-precision influencer discovery system built specifically for the Pakistani market. Two-phase pipeline: scrapes Instagram and TikTok via hashtag mining, commenter extraction, and seed graph traversal to discover nano and micro-influencers, then enriches each profile with real engagement metrics from the latest 15 posts. Filters bot accounts automatically. Includes a local analytics dashboard.,};2022 - 2026
Bachelor of Computer Science
Iqra University - Karachi
2019 - 2022
Diploma In CIT
Govt PolyTechnique Institute - Karachi
2009 - 2019
Secondary School Certificate
Bright Career School - Karachi
5 months ago
0
Building an Enterprise-Grade AI Voice Agent with Twilio, Deepgram, and Groq Llama-3.3 (Real-Time Telephony Automation)
4 Min Read
Building real-time AI voice agents over actual phone calls is one of the hardest engineering problems...
5 months ago
0
Building an AI-Powered Event Feedback System: Gemini 2.5 + FastAPI + Supabase + Automated PDF Reports
4 Min Read
After running an event, the feedback collection process is almost always broken. You get a...
5 months ago
0
I Built a Full-Stack Page Generation Engine with FastAPI + Next.js (And Here's the Architecture)
4 Min Read
There's a common problem in modern web development: you have structured data — JSON from a CMS, a...
1 years ago
0
How I Built a Facial-Expression Recognition Model with PyTorch (FER-2013, 72% Val Acc)
2 Min Read
TL;DR I trained a 3-block CNN in PyTorch on the FER-2013 dataset to classify 7 emotions....
© Developer Portfolio by Sameer Shah