Aksh Chaudhary

Developed an Automated Booking system reducing processing time by 99.98% (13 mins to less than 30 secs) using Flask, Celery (with Gevent), Azure Redis, CosmosDB, and Document Intelligence for OCR. Integrated Azure Pipelines for CI/CD with pytest, JWT for token validation, and Redis for token management. Implemented real-time task status polling, Salesforce/SharePoint integration, and ChatGPT 4o for attribute extraction. Set up log analytics with KQL and Azure Functions to feed PowerBI dashboards. Enabled zero-downtime deployment via staging slots. Also developed a Trade Compliance Chatbot using Databricks, GPT-4o-mini, Azure Blob, CosmosDB (vector DB), and secured all services via private endpoints in a VNet. Automated weekly document updates, logging, and PowerBI reporting via Azure Functions, reducing customer wait time by 100%. Working on fine-tuning gpt 4.1-nano for automated booking usecase.

• Worked on developing an Agentic RAG Chatbot for a hospital chain and deployed it using Fast API using WGSI server on an Azure Web App Service integrating CI/CD and enabling cross origin resource sharing and boosted the sales of the hospital chain upto 70% by giving personalised recommendation of health care services to the customers. • Worked on a Graph RAG which consisted of scraped data of around 1000+ diseases and has around 9+ millions entity nodes and 14+ million entity relationships and deployed it on a neo4j instance, this solutions is used to perform diagnosis of users. • Worked on ASR models to replace the existing whisper dependency to enable free of cost transcription services, deployed the models on VM and created endpoints for speech transcription. • Worked on speech synthesis to create speech from text using open source models like BARK. • Worked on voice cloning leveraging frameworks like BARK to generate a catalogue of voice clones for famous internet personalities. • Worked on an image generation from text use-case to generate images that were displayed in blogs of Tap Health’s site. • Worked on text-to-text translation models and replaced the Azure language translation endpoints(paid) with the open source model endpoints(free). • Worked on a scalable AI agent application that can be used to interview thousands of applicants at the same time autonomously without human intervention. • Used VAD for detecting periods of silence in the speech and Open AI whisper ASR for fluency and transcription. • Used various pdf parser tools like pdfplumber to parse candidate resumes. • Used complex scoring algorithms to weigh different sections, eg. fluency, resume, transcription, relevance, etc. • Used mediapipe to detect body language of candidates and assign a body language score. • Used GCP to deploy the AI backend using FastAPI router. • Used google cloud buckets to store candidate information.