mbasim78

Architected and deployed AI-driven voice intake, patient monitoring, and automated compliance agents for enterprise law and healthcare clients. Integrated Deepgram STT/TTS with OpenAI LLMs to ensure high-fidelity transcription accuracy, intent extraction, and context-aware conversations. Engineered a self-hosted LiveKit server on AWS, strategically configuring SIP/Twilio APIs for low-latency voice streaming and seamless call automation. Eliminated onboarding time by 40%, improved call accuracy by 35%, and designed modular structures to allow non-technical staff to execute workflow customizations.

Developed complex multi-agent pipelines utilizing Deepgram and OpenAI for real-time transcription and text-to-speech deployments. Deployed comprehensive intake agents featuring 8 integrated tools to completely automate client pipelines, scaling up to real-time appointment scheduling and rescheduling. Deployed backend LiveKit servers and Twilio integrations to securely support massive law firm calling architectures.

Engineered the strategic migration of LLM models from GPT-4 to LLaMA for medical summary generation pipelines, achieving a 25% cost reduction while maintaining strict system flexibility. Deployed models onto RunPod infrastructure to guarantee 99.9% uptime and handle the heavy scalability requirements of biomedical systems. Built and validated ML pipelines (ANN, SVM, Decision Trees) across 20+ medical datasets, successfully increasing predictive accuracy by 23%. Developed custom NLP workflows for clinical transcripts, achieving a 65% reduction in reading time for physicians and accelerating real-time decision-making by 20%.