Paarangat Rai Sharma

About

Founding-engineer-grade AI product builder who shipped 8 products in 6 months across production, beta, and pre-launch platforms (iOS, web, desktop, browser). Founding Engineer at JettyAI building GutBut, an AI-powered gut-health companion, with multi-codebase delivery and Claude agent + real-time streaming experience. Previously built real-time multi-model LLM orchestration with sub-3s latency at NULLZEC (HyberX) and supported healthcare AI workflows and compliance at Docwise. Also delivered multiple consumer iOS apps using Claude Code-driven agentic workflows at Valtrium. CGPA 9.54/10; Dean's List; 3x hackathon winner.

Experience

• Founding Engineer · JettyAIPresent

  Solo build of GutBut: Claude-powered AI gut-health companion · TestFlight closed beta with real users • Architected and delivered four codebases (~25k LOC TypeScript) in one month using Claude Code: NestJS backend, Next.js web with BFF, Expo iOS app, and Next.js superadmin for prompt management and LLM observability. • Designed the core Claude agent: 5 function-calling tools with a 2-turn cap, two-block system prompt with Anthropic ephemeral caching, and a preload-and-suppress optimizer that drops the tools schema from turn 0 — saves ~400 input tokens per common-path call. • Built a real-time Claude streaming experience for the iOS app: SSE with metadata interleave parsed via delimiter-aware look-behind buffer, React Native + Hermes streaming via expo/fetch, custom regex markdown parser, and a Reanimated streaming cursor. • Shipped a versioned prompt-management workflow with diff-gated publish (draft / published / archived states, side-by-side playground, atomic Postgres RPC); instrumented production LLM observability via PostHog $ai_generation events. Cut nightly insight- regen cost to near-zero with a SHA-256 hash-dedup cron.

• Foundational Engineer (Intern) (listed as Full-Stack Developer (Intern) in document) · Docwise2025

  Ambient AI scribe for NHS doctors (GDPR / UKCA / ICO / DTAC / DSPT compliant) • Shipping across four codebases (Chrome extension, dashboard, marketing site, NestJS backend) — 178 commits in 10 weeks working directly under the CTO on a healthcare AI product used by GPs, consultants, and junior doctors. • Built the Docwise Chrome extension from scratch in two weeks (Manifest V3, offscreen documents for tab audio capture, Side Panel API, IndexedDB) including patient consent flow and contextual fields transmitted alongside audio for AI note generation. • Owned the clinical recording workflow: Canvas-based real-time audio visualizer with FFT, useNavigationGuard preventing data loss across browser back / route changes / tab close; fixed an infinite-rerender bug by splitting a monolithic provider into memoized sub- contexts. • Built role-based personalization for three NHS personas (GPs, Consultants, Junior Doctors) and the Docwise marketing website from scratch (10+ pages with dynamic [slug] case studies, NHS compliance page) — including a one-week pixel-level Figma redesign sprint across ~40 components. Delivered full-stack features end-to-end including recycle bin and tutorial videos system (NestJS module + Supabase migration + frontend).

• Software Engineering Intern · NULLZEC (HyberX)2026 – 2026

  Real-time AI desktop application (Electron) · Multi-model LLM orchestration • Architected real-time pipeline (screen capture → dHash perceptual hashing → stability filtering → semantic dedup → multi-model LLM orchestration → streamed output) achieving sub-3s end-to-end latency. Designed gatekeeper + reasoning-model routing with speculative parallel execution, circuit breakers with adaptive throttling, 5-level JSON repair, 3-tier self-healing performance modes. Built the full Electron app (frameless overlay, Zod-validated IPC across 40+ channels, macOS notarization); decomposed a 2,100-line orchestrator into 8 DI modules with 304 Vitest tests.

• React Native Engineer (Intern) · Valtrium2026

  iOS developer using Claude Code-driven agentic workflows • Owned end-to-end React Native delivery for 6 consumer iOS apps in three months, with 736+ downloads across live apps. Live on App Store: Siddhartha (Buddhist storytelling), Kathavani (Hindu storytelling). Built and pending launch: Phoenix (21-day habit tracker), WatchIQ (luxury watch scanner), PokeScan (Pokémon card scanner), wake-up call app replacing alarms with phone calls. Also built internal content pipelines and analytics dashboards.

Education

• Bennett UniversityB.Tech, B.Tech CS (AI)

Similar talent on Pangea

Emanuel C.Senior Product Engineer (Vertical SaaS & AI Agents) | End-to-End Founder EngineerBuenos Aires, ArgentinaArjan A.An experienced full-stack developer adept at optimizing backend systems, and developing dynamic and user-friendly web interfaces.Toronto, ONJJamal M.Senior Software and AI Engineer | LLM Systems, RAG, Autonomous AgentsDetroit, MIHatim E.Full-Stack Engineer | Next.js · Supabase · AI Integrations · iOS — 6+ years shipping production MVPsSan Pablo City, Laguna