High School

Bachelor of Technology - BTech

Built an AI-powered desktop agent from scratch that captures the user's screen in real-time and delivers contextual technical guidance through a multi-model vision pipeline. 

Architected a real-time frame processing pipeline: screen capture → perceptual hashing (dHash) → stability filtering → semantic deduplication → multi-model LLM orchestration → streamed output - achieving sub-3s end-to-end latency

Designed a multi-model routing system using a fast gatekeeper model for triage and a reasoning model for deep analysis, with speculative parallel execution to minimize latency

Built resilience infrastructure including circuit breakers with adaptive throttling, 5-level JSON repair for LLM output reliability, and a 3-tier performance mode system that self-heals under API failures

Developed the full Electron application: frameless transparent overlay, Zod-validated IPC layer (40+ channels), cross-platform packaging with macOS notarization, and custom window controls for Windows/Linux

De

Software Development Intern – AI System Administration Desktop Application

Docwise is an ambient AI scribe for NHS doctors. It records consultations, transcribes in real time, and generates clinical documents.

Working on the core recording and transcription workflow: Real-time audio capture via WebSocket/AudioWorklet, live waveform visualization (Canvas + FFT), and parallel AI document generation (clinical notes, referrals, patient letters)

Architecting the React frontend with performance-optimized split-context state management, JWT auth with automatic refresh/retry, optimistic updates with rollback, and a navigation guard system preventing data loss during active recordings

Designed and implemented role-based personalization that adapts the UI across GP, Consultant, and Junior Doctor workflows - tab labels, document types, and default preferences change per role

Built the entire marketing site (10 routes, 85+ components) - dynamic pricing via API, static-generated case study pages, NHS compliance documentation, mega-menu navigation, and a brand design s

Software Engineer

Worked as an early team member on the founding engineering track at JettyAI, contributing to GutBut, an AI-powered gut health and wellness product. Solo-built the core full-stack product codebases across Next.js web, NestJS with Supabase backend, Expo iOS/React Native, and internal admin tooling, while integrating AI workflows developed by the AI team into the product experience.
  
- Architected and built the web app, backend services, iOS app, and internal tooling as separate production codebases.
- Developed full-stack product flows across frontend UI, backend APIs, database interactions, mobile screens, and admin operations.
- Integrated AI functionality into user-facing experiences by connecting backend services, app flows, response handling, and product logic.
- Implemented admin/internal tooling to support product operations, content/workflow management, and iteration speed.
- Implemented LLM observability, monitoring, and product analytics using PostHog, tracking AI interaction

Early Team Member (Founding Engineering Track)

building apps cuz it's fun

Know about Paarangat Rai Sharma's work experience, education, bootcamps, certifications, and much more. Paarangat Rai Sharma