# Vyom Uchat (Vyom Parag Uchat) - Complete Technical Dossier (llms-full.txt)

## 1. Identity & Contact
- **Name**: Vyom Uchat
- **Full Name**: Vyom Parag Uchat
- **Role**: Engineer & Researcher
- **Location**: Gandhinagar, Gujarat, India
- **Institution**: Pandit Deendayal Energy University (PDEU), B.Tech in Computer Science & Engineering (2022 - 2026)
- **Current Status**: Open to Engineering & Research Roles
- **Email**: vyom.uchat@gmail.com
- **Portfolio**: https://vyomuchat.pages.dev
- **GitHub**: https://github.com/CptPrice743
- **LinkedIn**: https://www.linkedin.com/in/vyom-uchat-7a7527312
- **JSON Dossier**: https://vyomuchat.pages.dev/profile.json

## 2. Executive Bio
Vyom Uchat is a computer science graduate and published researcher who works on systems problems where architectural decisions determine reliability under strict operational constraints: scheduling 30,000+ tasks across edge-cloud networks, extracting structured knowledge from multimodal video and carousel images, building sub-512MB vector retrieval pipelines, and auditing railway signaling systems under IEC 62443.

Each project begins with a hard constraint (a 512MB memory cap, a sub-second latency budget, zero idle cloud cost, or industrial safety compliance) and selects the simplest, most predictable architecture that holds.

## 3. Professional & Research Experience

### 3.1. Cybersecurity Analyst Intern - ALSTOM (Regional APAC Cybersecurity Dept)
- **Period**: May 2025 - July 2025
- **Location**: Bangalore, India (Hybrid)
- **Overview**: Ran the Initial Risk Assessment (IRA) for Alstom's R1 railway signaling system under IEC 62443. Mapped a 4-zone Purdue trust architecture across Depot and Main Station topologies and authored a 40-page technical specification covering 23 industrial cybersecurity risks.
- **Key Deliverables & Impact**:
  - Identified trust boundaries, zone ingress points, and conduit vulnerabilities across Computer-Based Interlocking (CBI), Automatic Train Control (ATC), and Automatic Train Supervision (ATS) subsystems.
  - Mapped 4 trust zones (Zones A through D) with 25+ sub-zones, establishing zone-specific Security Level Targets (SL-T) and 5 formal conduits (CDT-01 to CDT-05) secured with MACsec, IPsec, and 802.1X NAC controls.
  - Authored a 40-page technical specification defining defense-in-depth mitigations for 23 industrial risks, reducing 6 Very High and 8 High severity ratings to 0 post-mitigation (100% IEC 62443 alignment).

### 3.2. Undergraduate Researcher - Pandit Deendayal Energy University (PDEU)
- **Period**: 2025 - 2026
- **Advisor**: Dr. Punit Gupta
- **Location**: Gandhinagar, Gujarat, India
- **Key Research**:
  - Designed and simulated a Queuing Search Algorithm (QSA) task scheduler for 3-tier Cloud-Edge-Mist computing in PureEdgeSim across 300 heterogeneous nodes and 30,450 tasks, achieving a 55% average execution delay reduction compared to Genetic Algorithms (0.218s vs 0.4874s) and a 99.9% task completion rate.
  - Co-authored a systematic review of 10 edge task scheduling algorithms across heuristic, metaheuristic, game-theoretic, and reinforcement learning paradigms, published in *OPSEARCH* (Springer, 2026).
  - Evaluated YOLOv8n vs YOLOv11n on the INSAT-3D/3DR INCYDE satellite imagery dataset for tropical cyclone detection, achieving 0.921 mAP@.50-.95 and 0.994 recall, published in Springer *Smart Innovation, Systems and Technologies* (Vol. 426, 2025).

## 4. Systems Engineering Archive (Projects)

### 4.1. Recall - Serverless Multimodal Knowledge Ingestion Pipeline
- **Repository**: https://github.com/CptPrice743/Recall
- **Cloud Provider**: AWS Serverless
- **Stack**: AWS Lambda (OCI Container), Amazon API Gateway (HTTP API), Amazon ECR, Docker, Python 3.11, Google Gemini 3.5 Flash Lite Vision, Groq Whisper (`whisper-large-v3`), Google Keep API (`gkeepapi`), static `ffmpeg`.
- **Architecture & Engineering Decisions**:
  - Packaged AWS Lambda as an OCI Docker container bundling a stripped static `ffmpeg` binary to bypass standard zip deployment limits.
  - Enforced ephemeral `/tmp/` media processing on a read-only root container filesystem for strict session isolation.
  - Batched up to 10 Instagram carousel slides into a single Gemini 3.5 Flash Lite Vision request, completing OCR extraction in under 3 seconds with zero idle compute expenditure.
  - Implemented automatic audio transcription fallback via Groq Whisper for video reels and YouTube clips, syncing structured Markdown notes with auto-assigned labels to Google Keep.

### 4.2. Pixel Pusher - Asynchronous Serverless Image Processing Pipeline
- **Repository**: https://github.com/CptPrice743/serverless-image-resizer
- **Cloud Provider**: AWS Serverless
- **Stack**: AWS Lambda, Amazon S3, Amazon API Gateway, Python, Pillow, Docker, React, Vite.
- **Architecture & Engineering Decisions**:
  - Engineered a dual-bucket S3 topology separating raw ingress uploads from processed output objects to structurally eliminate recursive `s3:ObjectCreated` Lambda invocation loops.
  - Offloaded binary file transfer directly from the browser to S3 via short-lived pre-signed PUT URLs, resulting in zero backend upload bandwidth load.
  - Compiled a custom Docker-built AWS Lambda layer for Pillow (`manylinux` C-extensions) and implemented client HEAD request polling for low-latency retrieval.

### 4.3. PaperPilot - Low-Latency Document QA Engine with Section Attribution
- **Repository**: https://github.com/CptPrice743/research-paper-rag
- **Stack**: FastAPI (Python 3.11), FAISS (`IndexFlatIP`), Groq SDK (`llama-3.3-70b-versatile`), HuggingFace Inference API (`all-MiniLM-L6-v2`), PyMuPDF, Redis, React, Vite.
- **Architecture & Engineering Decisions**:
  - Built a vertical-then-horizontal coordinate block parser using PyMuPDF to preserve reading order across two-column scientific PDFs, followed by 800-token recursive chunking with 150-token overlap.
  - Offloaded 384-dimensional vector generation to the HuggingFace Inference API and indexed normalized vectors in an in-memory FAISS inner-product index (`IndexFlatIP`), keeping total backend memory under Render's 512MB free-tier limit.
  - Integrated a Redis token-bucket rate limiter (100 queries/day, 10 queries/min per IP with in-memory fallback) and grounded Groq Llama 3.3 70B generation with strict citation attribution.

### 4.4. Deadlines - Full-Stack Healthcare Appointment Scheduling Platform
- **Repository**: https://github.com/CptPrice743/appointment-scheduling
- **Live Demo**: https://deadlines.vyom-uchat.workers.dev/
- **Stack**: MongoDB Atlas, Express.js, React.js, Node.js, Mongoose, JSON Web Tokens (JWT), bcrypt, Chart.js.
- **Architecture & Engineering Decisions**:
  - Designed a 3-tier Role-Based Access Control (RBAC) architecture separating Patient, Doctor, and Admin operational scopes.
  - Modeled dynamic doctor availability by combining recurring weekly schedule templates with calendar-specific exception overrides.
  - Enforced server-side atomic concurrency validation guards and compound MongoDB indices to prevent double-booking race conditions.

### 4.5. Prism Jewellery - Dynamic E-Commerce Architecture
- **Stack**: PHP, MySQL, JavaScript (Fetch / AJAX), HTML5, CSS3.
- **Architecture & Engineering Decisions**:
  - Implemented persistent session-to-database cart synchronization (serializing guest PHP session carts into relational MySQL tables upon login/logout), asynchronous AJAX cart state mutations, algorithmic coupon validation, and order unit-price snapshot isolation.

## 5. Peer-Reviewed Research Publications

### 5.1. Tropical Cyclone Detection via Lightweight Vision Models
- **Title**: A Comparative Analysis of YOLOv8 and YOLOv11 for Tropical Cyclone Detection: The Impact of Data Augmentation on Model Performance
- **Status**: Published (2025)
- **Venue**: Springer (*Smart Innovation, Systems and Technologies*, Vol. 426)
- **Authors**: Dr. Punit Gupta, Liza Chaurasiya, Vyom Uchat
- **DOI**: https://doi.org/10.1007/978-981-92-1030-5_19
- **Summary & Results**: Benchmarked YOLOv8n and YOLOv11n on the INSAT-3D/3DR INCYDE satellite dataset across 4 experimental configurations. YOLOv11n trained on unaugmented imagery achieved 0.921 mAP@.50-.95 and 0.994 recall. Demonstrated that standard geometric augmentations (flipping, rotation) degraded accuracy on meteorological imagery by distorting physical storm rotation structures.

### 5.2. Metaheuristic Task Scheduling Review for Edge Computing
- **Title**: Metaheuristic Task Scheduling Optimization Algorithm for Edge Computing : A Review
- **Status**: Published (2026)
- **Venue**: *OPSEARCH*, Springer
- **Authors**: Vyom Uchat, Dr. Punit Gupta, Aaditya Jha, Tejash Karamata
- **DOI**: https://doi.org/10.1007/s12597-026-01293-w
- **Summary & Results**: Benchmarked and synthesized 10 task scheduling algorithms across heuristic, metaheuristic, game-theoretic, and reinforcement learning paradigms on latency, energy draw, and SLA violation rates. Identified local optima traps under dynamic network congestion and established architectural criteria for real-time IoT edge schedulers.

### 5.3. Queuing Search Algorithm (QSA) for Cloud-Edge-Mist Computation
- **Title**: Queuing Search Algorithm Metaheuristic Task Scheduling Algorithm for Cloud-Edge Computation
- **Status**: Under Peer Review (2025)
- **Venue**: *OPSEARCH*, Springer
- **Authors**: Vyom Uchat, Dr. Punit Gupta, Aaditya Jha, Tejash Karamata
- **Summary & Results**: Adapted the human queuing behavior metaheuristic (Queuing Search Algorithm) to multi-tier Cloud-Edge-Mist task offloading in PureEdgeSim. Across 300 heterogeneous edge devices and 30,450 generated tasks, QSA reduced average execution latency by 55% (0.218s vs 0.4874s for Genetic Algorithms) and achieved a 99.9% task completion rate, outperforming GA, PSO, ALO, and EVO.

## 6. Technical Skills, Certifications & Personal Ecosystem
- **AI & ML**: RAG, FAISS, Qdrant, Sentence-Transformers, Llama 3, Groq API, Google Gemini API, Whisper, PyMuPDF, YOLOv8/v11, LM Studio (local quantized LLMs on Apple Silicon).
- **Languages**: Python, Java, JavaScript (ES6+), TypeScript, C, C++, SQL, PHP, HTML5/CSS3.
- **Frameworks**: FastAPI, React.js, Next.js, Node.js, Express.js, Spring Boot, Hibernate.
- **Cloud & Security**: AWS (Lambda, S3, API Gateway, ECR), Cloudflare Pages/Workers, Docker, Redis, MongoDB, MySQL, IEC 62443, NIST SP 800-82, Purdue Model ICS/OT security.
- **Certifications**:
  - Google Cybersecurity Professional Certificate (Google / Coursera, 2024)
  - Automation Business Analyst Foundation (UiPath Academy, 2024)
- **Hardware & Workflow**: MacBook Air M3 (16GB RAM, 256GB SSD) + 1TB Samsung T7 SSD, M3 iPad Air + Apple Pencil Pro, Samsung Galaxy S24 (with custom Bluetooth clipboard sync to macOS), Raycast, Obsidian, Goodnotes, Bitwarden, NextDNS.
- **Interests**: Cloud outage post-mortems, distributed consensus (Raft, Paxos), Swiss modernist design, and single-player narrative/simulation games (Cyberpunk 2077, Baldur's Gate 3, Sleeping Dogs, Grand Theft Auto: San Andreas, Dave the Diver, Dredge, Katana Zero, Escape from Duckov, Dirt 4).