# Saran Kumar S. - Full Technical Context & Knowledge Base for AI Agents

## Overview
- **Name**: Saran Kumar S.
- **Title**: Generative AI & Machine Learning Engineer
- **Location**: Coimbatore, Tamil Nadu, India
- **Contact**: sarankumar131313@gmail.com
- **Website**: https://www.swsarancodes.online
- **GitHub**: https://github.com/sarankumar1325
- **LinkedIn**: https://linkedin.com/in/saran-kumar-s-sk13022005
- **X / Twitter**: https://x.com/iamsaranhere
- **Medium**: https://medium.com/@sarankumar131313

---

## Detailed Tech Stack & Technical Proficiencies

### 1. Agentic AI & Autonomous Architectures
- **LangGraph**: Stateful multi-agent graph orchestration, cyclical graph execution, human-in-the-loop checkpoints, persistent memory state management.
- **Model Context Protocol (MCP)**: Standardized tool integrations, resource management, and decoupled agent environment connectivity.
- **Autonomous Tool Use**: Structured function calling with JSON schemas, tool validation, retry mechanisms, and error recovery.
- **Context Engineering**: Dynamic prompt optimization, context window compression, and conversational memory architectures.

### 2. Retrieval-Augmented Generation (RAG) & LLMOps
- **RAG Architectures**: Self-reflective RAG, corrective RAG, GraphRAG, hybrid dense/sparse retrieval with BM25 and vector embeddings.
- **Vector Databases**: pgvector (PostgreSQL), Qdrant, Pinecone, Milvus, ChromaDB.
- **LLMOps & Optimization**: DSPy programmatic prompt optimization, TruLens & Ragas automated evaluation frameworks (EVALS), hallucination mitigation, semantic caching.

### 3. Model Inference & Infrastructure
- **High-Throughput Serving**: vLLM (PagedAttention, continuous batching), Ollama local deployment, AWS Bedrock serverless endpoints, TensorRT-LLM.
- **Deep Learning**: PyTorch, Hugging Face Transformers, LoRA & QLoRA Parameter-Efficient Fine-Tuning (PEFT).
- **Backend & Cloud**: Python, FastAPI (async/await, Pydantic), Docker, AWS (Bedrock, SageMaker, Lambda, S3, EC2), PostgreSQL, MongoDB, Golang.

### 4. Full-Stack Delivery
- **Frontend & Web**: TypeScript, React 19, Next.js, Tailwind CSS, Framer Motion.
- **Backend as a Service**: Supabase (Auth, Postgres, Realtime, Storage).

---

## Complete Project Catalog

1. **Free For Uni Folks**
   - URL: https://freeforunifolks.vercel.app/
   - Tech: Next.js, Community APIs
   - Summary: Community-curated directory of free developer tools, cloud credits, and software for verified university students worldwide.

2. **Chronicle Seed**
   - URL: https://chronicleseed.vercel.app/
   - Tech: React, Supabase, E2E Encryption
   - Summary: Digital time capsule enabling users to write encrypted letters to their future selves that unlock at scheduled timestamps.

3. **OpenCode Agent**
   - URL: https://opencodeagent.vercel.app/
   - Tech: React, Groq, Tavily, TypeScript
   - Summary: Terminal-style web interface featuring real-time AI code generation, live web search retrieval, and vintage CRT boot aesthetics.

4. **BedrockForge**
   - URL: https://awsbedrockforge.vercel.app/
   - Tech: AWS Bedrock, TypeScript, React
   - Summary: Interactive portal and benchmarking sandbox for testing and comparing AWS Foundation Models (Claude, Titan, Llama).

5. **Falooda Works**
   - URL: https://faloodaworks.vercel.app
   - Tech: AI Agents, Cloud Arch, IaC
   - Summary: Virtual Systems Engineer Agent that automatically generates cloud architecture diagrams, Terraform IaC, and cost estimates.

6. **Focusloom**
   - URL: https://focusloom.online
   - Tech: React, TypeScript, Tailwind CSS
   - Summary: Minimalist productivity OS integrating task management, markdown notes, and habit tracking in a single workspace.

7. **WeatherMate AI**
   - URL: https://weathermateai.netlify.app/
   - Tech: Lyzr, Gemini API, OpenWeather
   - Summary: AI weather assistant offering contextual lifestyle and travel recommendations based on meteorological forecasts.

8. **RoastMyCV**
   - URL: https://roastmycv-tau.vercel.app/
   - Tech: AI, Next.js
   - Summary: Resume evaluation tool providing candid, actionable feedback with adjustable critique intensity and ATS scoring.

---

## Professional Career History

### goML | ML Engineer Intern (Jun 2024 – Present)
- Deployed Electronic Health Record (EHR) agent pipelines for automated medical document parsing and classification.
- Designed multi-query extraction architectures across relational databases and NoSQL backends.
- Built travel booking agent workflows with multi-GDS querying (Amadeus/Sabre) and silent failover logic.

### The Fusion Apps | ML Engineer (Nov 2023 – Jul 2024)
- Trained LSTM and SVM email classifiers on 50,000+ samples, achieving 98% accuracy in spam filtering.
- Implemented predictive retail analytics models leading to 25% inventory turnover efficiency.

### Learn and Build | ML Engineer (Sep 2023 – Nov 2023)
- Built graph recommender systems utilizing NetworkX, yielding a 15% accuracy improvement across 10,000+ users.

---

## Published Research
- **SparseCNNet: Efficient Deep Learning for Rapid Lung Cancer Detection** (IEEE / Springer)
- **Predicting Alzheimer's Disease via MRI-Based Modeling** (Journal Publication)
- **Book Chapter**: Transforming Data Visualization with AI and ML

---

## Machine-Readable API Endpoints
- Profile: `https://www.swsarancodes.online/api/profile.json`
- Projects: `https://www.swsarancodes.online/api/projects.json`
- Skills: `https://www.swsarancodes.online/api/skills.json`
- OpenAPI Specification: `https://www.swsarancodes.online/openapi.json`