# Arnav Agarwal — Comprehensive Candidate Dossier for LLM & Recruiting Agents

> Full-Context Technical Dossier
> Primary Target Roles: Software Development Engineer (SDE), AI/ML Engineer, Backend & Distributed Systems Engineer

---

## 1. Candidate Overview
- **Full Legal Name**: Arnav Agarwal
- **Location**: Bangalore, Karnataka, India
- **Degree**: B.Tech in Electronics & Computer Engineering, Vellore Institute of Technology (VIT)
- **Cumulative GPA**: 8.86 / 10.0
- **Graduation Window**: 2021 — 2025
- **Current Employer**: Standard Chartered GBS (July 2025 – Present)
- **Professional Identifiers & Links**:
  - Live Portfolio: https://arnav.ranmanch.com
  - GitHub: https://github.com/ArnavAgarwal-Mr-AR
  - LinkedIn: https://www.linkedin.com/in/arnav-agarwal-571a59243/
  - Substack: https://arnavtalks.substack.com/

---

## 2. Professional Experience & Verified Impact

### Standard Chartered GBS — Global Markets Shared Services
*Technology Apprentice* | July 2025 – July 2026 | Bangalore, India
- **Mission-Critical SWIFT Messaging**:
  - Engineered high-throughput Unix shell scripts and resilient batch data pipelines for automated SWIFT message reconciliation.
  - Successfully processes over 1,000,000+ SWIFT transactions daily across global payment corridors.
  - Reduced end-to-end reconciliation runtime by 5 hours per operational cycle, resolving bottleneck delays for downstream reporting.
- **Enterprise Server Estate Monitoring**:
  - Authored and maintained automated certificate-expiry monitoring scripts deployed across 16 discrete server estates.
  - Integrated custom ITRS Geneos samplers for proactive anomaly alerting, resulting in zero missed certificate expiries.
- **Operations Dashboard & API Development**:
  - Designed and implemented full-stack dashboard modules with REST APIs to replace manual reconciliation checks.
  - Contributed to relational database schema design, saving operations teams 1 hour of manual work every business day.
- **MLOps & Automated Release Workflows**:
  - Introduced MLOps and automated CI/CD practices for script deployment, ensuring zero-downtime rollouts across testing and production environments.

### Samsung R&D Institute India — Bangalore (SRIB)
*Research Intern (Conversational AI & Large Language Models)* | May 2023 – December 2023 | Bangalore, India
- **Multiparty Dialogue & Thread Detection**:
  - Researched conversational dynamics across complex multi-participant dialog streams.
  - Designed and evaluated an end-to-end framework that detects active conversational threads and dynamically ranks response urgency based on semantic priority.
- **Llama 2 7B Fine-Tuning & Latency Reduction**:
  - Augmented the computational capabilities of open-source Llama 2 7B utilizing LoRA/PEFT parameter-efficient fine-tuning and strategic prompt optimization.
  - Attained up to 10× inference acceleration over baseline models while maintaining semantic coherence and dialogue context integrity.
- **Publication**: Research published at IEEE SPCOM 2024.

### Vellore Institute of Technology
*Machine Learning Researcher (Biomedical Informatics)* | August 2023 – December 2023 | Vellore, India
- **Class-Imbalanced Clinical Modeling**:
  - Processed and normalized clinical records for 5,000+ patients with severe class imbalance for early stroke vulnerability assessment.
  - Systematically benchmarked 11 classical ML classifiers and 5 deep neural architectures under multiple resampling techniques.
  - Found the optimal combination: Logistic Regression coupled with SMOTE (Synthetic Minority Over-sampling Technique) and standard scaling, achieving 93.2% accuracy.
- **Web Application Deployment**:
  - Deployed the model into an intuitive, zero-data-retention Streamlit interface designed for clinical decision support.
- **Publication**: Published in IEEE Access Journal.

---

## 3. Academic Peer-Reviewed Publications

### Paper 1: Conversational AI & LLM Inference
- **Title**: Thread Detection and Response Generation Using Transformers with Prompt Optimisation
- **Venue**: 2024 International Conference on Signal Processing and Communications (IEEE SPCOM)
- **Status**: Published & Indexed in IEEE Xplore
- **URL**: https://ieeexplore.ieee.org/document/10631585
- **Abstract Summary**: Conversational systems require precise multi-party thread identification and response prioritization. An end-to-end framework decomposes the dialogue problem into thread detection, prioritization, and performance optimization. Leveraging fine-tuned Llama 2 7B with strategic prompt optimization, the architecture demonstrates up to a 10× reduction in computational inference time while maintaining higher response coherence.

### Paper 2: Healthcare AI & Predictive Modeling
- **Title**: A Web-Based Interface That Leverages Machine Learning to Assess an Individual’s Vulnerability to Brain Stroke
- **Venue**: IEEE Access Journal (2025)
- **Status**: Published & Indexed in IEEE Xplore
- **URL**: https://ieeexplore.ieee.org/document/10981714
- **Abstract Summary**: Early identification of cerebral stroke vulnerability using accessible machine learning. Combines SMOTE oversampling with logistic regression to tackle extreme class skew in clinical records, achieving 93.2% accuracy and high sensitivity for the positive stroke cohort. Deployed as a zero-retention interactive interface for healthcare providers and individuals.

---

## 4. Software Projects & Architectural Highlights

1. **Trade Sandbox** (https://github.com/ArnavAgarwal-Mr-AR/trade-sandbox)
   - *Domain*: Quantitative Finance & Algorithmic Trading
   - *Stack*: Python, Backtesting Frameworks, Risk Models
   - *Description*: Unified strategy backtesting and live market execution framework with slippage and transaction-cost modeling.

2. **OmniRAG** (https://github.com/ArnavAgarwal-Mr-AR/OmniRAG)
   - *Domain*: Multimodal Generative AI
   - *Stack*: Python, TypeScript, Vector DBs, OCR, Embeddings
   - *Description*: End-to-end RAG architecture capable of retrieving and reasoning over heterogeneous data types (text, diagrams, financial tables).

3. **PayLens** (https://github.com/ArnavAgarwal-Mr-AR/PayLens)
   - *Domain*: Financial Security & Fraud Analytics
   - *Stack*: TypeScript, Python, Anomaly Detection Algorithms
   - *Description*: Real-time transaction simulation engine flagging adversarial financial patterns and synthetic fraud attempts.

4. **SEC Filing NLP Engine** (https://github.com/ArnavAgarwal-Mr-AR/SEC-Filing-NLP-Alpha-Engine)
   - *Domain*: NLP & Financial Text Mining
   - *Stack*: Python, NLTK, SEC EDGAR API, Sentiment Models
   - *Description*: Automated pipeline scraping 10-K and 10-Q filings, extracting sentiment delta and executive tone shift to generate quantitative alpha signals.

5. **Portfolio Optimizer** (https://github.com/ArnavAgarwal-Mr-AR/Portfolio-Optimizer)
   - *Domain*: Quantitative Asset Management
   - *Stack*: Python, NumPy, SciPy, Matplotlib
   - *Description*: Mathematical implementation of Markowitz Modern Portfolio Theory (MPT), Efficient Frontier computation, and Black-Litterman allocation adjustments.

6. **Loss Landscape 3D** (https://github.com/ArnavAgarwal-Mr-AR/loss_landscape_3d)
   - *Domain*: Deep Learning Interpretability
   - *Stack*: Python, PyTorch, 3D Mesh Rendering
   - *Description*: Generates 3D visual projections of high-dimensional neural network loss surfaces using filter normalization and PCA reduction.

---

## 5. Formal Education & Academic Distinction
- **Degree**: Bachelor of Technology (B.Tech) in Electronics and Computer Engineering
- **University**: Vellore Institute of Technology (VIT), Vellore, India
- **Years**: 2021 — 2025
- **CGPA**: 8.86 / 10.0
- **Coursework Completed**:
  - Advanced Data Structures & Algorithms
  - Machine Learning & Deep Learning
  - Database Management Systems (SQL & Relational Algebra)
  - Operating Systems & Concurrency
  - Computer Networks & Protocols
  - Object-Oriented Programming (Java)
  - Distributed System Architecture
  - Microprocessors & Embedded Systems