AI Engineer Intern - EGeeks Global - Rawalpindi, Pakistan
(2025-06 - 2025-09)
- Collaborated within a cross-functional team of developers.
- Contributed to delivery of phase 1 of Viva-Bot, an enterprise-grade application that provides professional-level interview and thesis defense.
- Built and optimized React-based UI components for dynamic user interaction and real-time feedback.
- Integrated OpenAI API for mock interview generation and Deepgram API for real-time transcription during live interviews, enabling dynamic, context-aware feedback in Viva Bot.
- Researched open-source lip-sync and video generation models (MuseTalk, Wav2Lip, Wan2.1 and Hallo3).
- Worked on webRTC for real-time streaming. Configured streaming APIs HeyGen, ANAM, Tavus and D-ID.
- Integrated Retell AI agents with n8n workflows to automate outbound voice calls prompting customers to clear dues via a secure payment line.
- Designed schema-first Supabase integration, ensuring normalized customer data, transactional integrity, and secure access control.
- Designed a dynamic dashboard displaying customer details and payment status, real-time audio playback of call interactions, transcription logs for compliance and audit.
- Optimized agent workflows for concurrency and reliability, reducing manual intervention in debt recovery.
- Worked on a Flask based web application for managing outbound calls through VAPI.ai integration. The system includes user authentication, subscription management, and bulk call processing capabilities.
- Automated call scheduling, rescheduling, follow-up and timezone-aware logic for global customers.
- Integrated AI-driven transcript analysis for call outcome evaluation and dynamic workflow decisions.
Full Stack Automation Engineer - Evernect - Islamabad, Pakistan
(2025-09)
- Developed multilingual voice assistant (UK, FR, DE) with auto language detection and dynamic response routing.
- Engineered conversational logic with office hours validation, holiday detection, and context-aware dialogue flow.
- Implemented call routing with 10+ department protocols using caller type, query and geographic ZIP code analysis.
- Built automated technician lookup via postal code database for Germany and Austria, improving accuracy.
- Designed intelligent decision-tree architecture handling 5 main query categories (retailers, service requests, outlet inquiries, emergencies, internal departments) with nested conditional logic and personalized response paths.
- Engineered a flow with multi-objective optimization, achieving 95%+ contact capture and enquiry completion.
- Built dynamic pricing tier explanation logic for 4 subscription levels integrating real-time calendar availability via API.
- Implemented state management to track contact details, appointment preferences, and customer intent.
- Created fail-safe recovery and validation mechanisms ensuring critical information capture before call termination, reducing incomplete lead conversions.
- Designed natural language appointment scheduling with time-slot optimization (9 AM–5 PM) based on real-time calendar availability.
- Developed an outbound call system handling 1,500+ calls per campaign, reducing manual workload by 95% across UK and Swiss markets with location-aware responses and international regulatory compliance.
- Engineered retry mechanism re-attempting unanswered leads 3× daily for 2 days, optimizing contact rates and ensuring maximum lead coverage.
- Built scalable concurrent calling infrastructure supporting up to 10 simultaneous outbound calls while maintaining system stability and preventing concurrency limit violations through intelligent queue management.
- Implemented multi-stage call flow automation including IVR navigation, dynamic email verification with phonetic spelling confirmation, and real-time lead qualification across international markets.
- Created adaptive email handling with fallback, phonetic spelling, and multi‑step confirmation for 99%+ accuracy.
- Reduced average appointment scheduling time by 70% through intelligent automation.
- Achieved 95%+ accuracy in patient information collection through multi-step validation.
- Handled 100% of after-hours scheduling requests without human intervention.
- Decreased administrative staff workload by automating routine inquiry responses and scheduling tasks.
- Built patient profile management system with CRUD, validation, phonetic confirmation, and secure storage integrated with PatientQ CRM.
- Integrated payment and scheduling workflows with deposits, tiered pricing, insurance via Mentaya API, and automated reminders, reducing no-shows and boosting clinic efficiency.
- Developed an AI-powered quoting and invoicing assistant for tradespeople, enabling instant PDF quote generation via WhatsApp integration.
- Architected modular workflows to automate client communication, quote creation, and delivery, reducing manual effort by 80%.
- Integrated analytics dashboards to track quotes and optimize business performance, providing actionable insights.
- Implemented Stripe payment gateway with recurring subscription billing, supporting Trial, Basic, and Standard tiers with automated invoicing and upgrade/downgrade logic.
- Designed and managed database architecture using Supabase (PostgreSQL) with Prisma ORM, enabling scalable, type-safe queries and efficient data modeling for quotes, users, and subscriptions.
