Hire Akshan P. - Generative AI Engineer

Built an end-to-end clinical scribing system handling audio, documents, and medical images using FastAPI. Implemented speech-to-text pipelines with Faster-Whisper and Azure Speech SDK, including speaker diarization using pyannote. Designed temporal overlap and heuristic-based speaker attribution to distinguish Doctor vs Patient conversations. Generated structured clinical outputs (SOAP notes, chief complaints, triage insights) using LLM with strict JSON prompting. Applied medical image segmentation using MedSAM and vision-enabled LLMs for clinical image analysis and reporting. Architected an AI-Powered Conversational Health Assistant integrating Large Language Models (LLMs) for intent classification, medical question handling, and dynamic patient interaction. Deployed sophisticated dialogue management systems to process patient chief complaints, conduct structured symptom assessments, and generate LLM responses, enhancing user experience. Engineered data processing and retrieval mechanisms for medical history databases, enabling context-aware LLM interactions and automated report generation for clinical insights. Utilized GPT-4, Mixtral, and fine-tuned models for response generation. Integrated Whisper Large v3, GPT-4o-transcribe, and TTS HD for seamless end-to-end speech conversation. Developed a prototype healthcare interoperability platform leveraging Allscripts Unity APIs (Practice Management and EHR) to retrieve patient demographics, clinical records, and support appointment scheduling workflows. Converted unstructured data into FHIR R4 (US-Core 6.1.0) transaction bundles supporting Patient, Encounter, Observation, Condition, and Medication resources. Implemented terminology normalization using LOINC, ICD-10-CM, SNOMED CT, RxNorm, CPT, and UCUM for standards-compliant interoperability. Built voice-based encounter editing, clinical summarization, and report generation using STT and LLMs.