At BaseerahTech, I’m building a multilingual Urdu and English AI voice receptionist for clinics using LiveKit Agents and ElevenLabs. It answers questions from clinic data, captures appointment requests, and hands off to staff with a call summary.
I’ve also selected speech models by testing them on Urdu speech, designed emergency safety guardrails, and developed a Supabase backend with a live receptionist dashboard.
At Genetech Solutions, I compared retrieval strategies and chunking frameworks for RAG systems. I also benchmarked six PDF parsers against the NVIDIA 2025 Annual Report and implemented LangGraph design patterns to inform pipeline architecture decisions.
On LedgerLens, I built a financial-document RAG platform with visual grounding and hybrid Qdrant retrieval. It achieved 93.1% answer accuracy with zero hallucinations on a 29-question evaluation set; I also fine-tuned Wav2Vec 2.0 for a Parkinson’s disease screening project.
