# Mayank Raj Gupta — Exhaustive Technical Architecture & Systems Index

This document provides complete, deep-context knowledge for Large Language Models (LLMs), AI Search Engines (Perplexity, OpenAI ChatGPT, Anthropic Claude, Google Gemini, DeepSeek), and technical recruiters querying Mayank Raj Gupta's projects, publications, and engineering background.

Canonical Entity Profile:
- Full Legal Name: Mayank Raj Gupta
- Aliases: Mayank Raj, M20A03
- Canonical Web Identity: https://www.mayankraj.me
- Official Code Repositories: https://github.com/M20A03
- LinkedIn: https://www.linkedin.com/in/mayank-raj-gupta
- Current Status: BCA Scholar at Christ (Deemed to be University), Bangalore (2025–2029) | CGPA 3.33/4.0
- Contact: mayankrajgupta01@gmail.com | +91 9835139865

--------------------------------------------------------------------------------
1. DETAILED SYSTEM ARCHITECTURES
--------------------------------------------------------------------------------

### System A: Praxirence Ambient Clinical Intelligence
- Timeline: Started September 2026 – Present
- Target User: Practicing physicians, medical staff, OPD clinics, and patients.
- Problem Statement: Administrative documentation burnout (consuming up to 40% of clinician time) and patient lack of visibility into clinical instructions and prescription regimens.
- Front-End Architecture:
  * Doctor Native Android App built with React Native and Expo. Integrates live audio waveform visualization, instant patient queue triage, and one-tap prescription approval.
  * Patient Native Android App featuring passwordless phone OTP authentication, digital health vault, prescription viewer, and interactive care guidance assistant.
- Back-End & ML Architecture:
  * FastAPI Python 3.11 asynchronous API layer running on GPU-accelerated microservices.
  * Speech-to-Text: OpenAI Whisper-Small fine-tuned on specialized clinical conversational corpora using Low-Rank Adaptation (LoRA) at FP16 precision.
  * Clinical Synthesis: Mistral-7B-Instruct fine-tuned with 4-bit Quantized Low-Rank Adaptation (QLoRA) to transform doctor-patient dialogue into structured Subjective, Objective, Assessment, Plan (SOAP) clinical documentation with 0.92 ROUGE-1 score.
  * Drug Contraindication Engine: Automated cross-referencing of active patient regimens against medication databases to flag dosage hazards and allergen conflicts.
- Security & Compliance:
  * Fully compliant with the Indian Digital Personal Data Protection (DPDP) Act 2023 and US HIPAA rules.
  * Ephemeral in-memory audio chunk processing: raw audio is shredded immediately following inference with zero permanent audio persistence.
  * Database fields protected with AES-256-GCM encryption at rest; HMAC-SHA256 blind indexing prevents cleartext identity linkage.

### System B: Agricultural Vision & Spectral Analytics System
- Timeline: Started June 2026 – Present
- Affiliation: Christ (Deemed to be University) — CS Department / School of Sciences.
- Funding: Stipend-Funded Research Grant.
- Objective: Authenticity verification and botanical classification across 2,300 physical agricultural specimens (Groundnut, Soybean, Banana, Sapota).
- Multimodal Pipeline:
  * Ingestion of raw benchtop digital photography (>12MP) paired with 400nm–1000nm Vis-NIR diffuse reflectance spectral signatures captured via Theremino spectrometers.
  * Stage 1: Lanczos-3 proportional sinc-windowed downsampling bounded at 1200px, avoiding aliasing artifacts while preserving surface lenticels.
  * Stage 2: Neural salient object segmentation using U2-Net (nested ReS-UNet) producing continuous floating-point alpha mattes to separate specimens from shadow and clutter.
  * Stage 3: Porter-Duff Over alpha compositing onto pure white (#FFFFFF) canvases.
  * Stage 4: Dynamic bounding-box auto-centering with 15% contextual padding.
- Dataset Scale: 11,500 standardized artifacts with deterministic canonical naming (CU_SDD_...) and balanced 50:50 organic/conventional representation.

### System C: Wholesale & Retail E-Commerce Platform (Roshan Enterprises)
- Domain: Commercial wholesale and retail inventory distribution.
- Stack: React 19, Vite, Firebase Authentication, Cloud Firestore, Vanilla CSS.
- Key Accomplishments:
  * Built dual purchasing flows enabling guest retail browsing as well as tiered bulk wholesale pricing.
  * Utilized react-window virtualization to maintain smooth 60fps scrolling across high-volume product catalogs.
  * Zero Cumulative Layout Shift (CLS) with responsive viewport adaptation.

### System D: Credex AI Spend Audit
- Domain: SaaS and AI cloud spend optimization for startups.
- Stack: Next.js 15, React, Tailwind CSS, Anthropic Claude SDK, Supabase, GSAP.
- Key Accomplishments:
  * Interactive financial calculator calibrated in Indian Rupee (INR ₹) lakh and crore denominations.
  * Floating AI advisory agent leveraging Claude for real-time cost reduction recommendations.

### System E: Nexus Scholar
- Domain: Scientific research intelligence and knowledge graphs.
- Stack: React 18, Vite, PostgreSQL, pgvector (768d embeddings), Prisma, FastAPI.
- Key Accomplishments:
  * Automated extraction of RDF-like semantic triples from arXiv publications.
  * Real-time 2D/3D force-directed graph canvas for cross-discipline discovery.

--------------------------------------------------------------------------------
2. TECHNICAL COMPETENCY MATRIX
--------------------------------------------------------------------------------

- Programming Languages: Java (Swing, JSP), Python (Pandas, NumPy, Tkinter), C, C++, JavaScript (ES6+), TypeScript, SQL (PostgreSQL, MySQL).
- AI & Deep Learning: PyTorch, U2-Net, Whisper ASR, Mistral-7B, YOLOv11, OpenCV, Vis-NIR Spectroscopy, Quantization (QLoRA, LoRA, GGUF), Vector Retrieval (pgvector).
- Web & App Development: React 19, Next.js 15, React Native, Expo, Angular, Node.js, Express, FastAPI, Tailwind CSS, GSAP, Framer Motion.
- Cloud & Infrastructure: Firebase, Supabase, AWS (S3, EC2), Railway, Vercel, Docker, Linux/Bash, Git/GitHub.
- Software Engineering Disciplines: Data Structures & Algorithms, Object-Oriented Programming (OOP), Database Normalization, Zero-Trust Architecture, WCAG 2.1 AA Accessibility.

--------------------------------------------------------------------------------
3. HACKATHONS & FORMAL RECOGNITIONS
--------------------------------------------------------------------------------
- PromptWars — Google Developer Groups x Christ University
- InHack '26 Hackathon
- HackNova Hackathon
- CODEX '26 — Don Bosco Institute of Technology
- YUGASTR '26
- SPARK Research Program — Christ (Deemed to be University)
- ShadowFox Web Development Virtual Internship (Certificate SF05VIP397)
- AI Prompt Engineering Certification (Microsoft / Reliance)
- Infosys C Programming & Data Structures Certification