~/sparrow84001 (main) • Full Stack + AI EngineerOpen to RolesWebMCP • AEO Verified

Sayanta Neogi

Software Engineer • AI & RAG Architect • Cloud DevOps

Experienced Full Stack Developer and AI Engineer with 4+ years of expertise in building scalable web applications, implementing AI-powered solutions, and establishing DevOps practices. Proficient in modern frameworks including React, Laravel, Django, Flask, and cutting-edge AI technologies such as Amazon Bedrock, LangChain, Hugging Face, Model Context Protocol (MCP), and RAG architectures. Skilled in containerization with Docker, CI/CD pipelines, and cloud deployments across AWS, Azure, and GCP.

React & AstroPython / Django / FlaskAmazon Bedrock & pgvectorLangChain & MCP ServersDocker & Kubernetes
4+ YearsProduction Experience
24Public GitHub Repos
AWS & AzureCloud Deployments
Full Stack & AICore Paradigms
sparrow84001/mcp-seo • TypeScript
// sparrow84001/mcp-seo • AI-Powered SEO & Growth Audit MCP Server
import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
import { z } from "zod";
import { discoverProject } from "./analyzer/discovery.ts";

export const server = new McpServer({
  name: "mcp-seo",
  version: "1.0.4"
});

// Register tool: Framework & Route Discovery
server.registerTool(
  "seo_discover_project",
  {
    description: "Discovers framework (Astro, Next.js, Laravel), routes & metadata.",
    inputSchema: { projectPath: z.string().describe("Project root directory") }
  },
  async ({ projectPath }) => {
    const discovery = await discoverProject(projectPath || ".");
    return { content: [{ type: "text", text: JSON.stringify(discovery, null, 2) }] };
  }
);
Active: McpServer v1.0.4 • Runtime: Bun
sparrow84001/mcp-seo
Interactive Tool Sandbox • Model Context Protocol

Live MCP & RAG Playground

Execute live simulations of the authentic Model Context Protocol tools and RAG search pipelines Sayanta engineered. Inspect real AST analysis, framework discovery, and vector embeddings.

Source:src/analyzer/discovery.ts

Discovers web framework (Astro, Next.js, Laravel), routes & static metadata from project root AST.

Tool Input Schema (Zod)JSON Payload
{
  "projectPath": "d:/SNtemp/astro/SN_portfolio",
  "detectSitemap": true,
  "scanRoutes": true
}
Runtime: Bun v1.4Transport: stdio / SSE
Tool Response
Sandbox ReadyClick "Execute Tool" to dispatch simulated MCP payload and inspect live AST and vector search results.
AST Validator ActiveJSON-RPC 2.0 Compliant
Distributed System Architecture

Enterprise AI & AST Pipeline Topology

Interactive schematic of Sayanta’s production architecture connecting LLM agent clients with Model Context Protocol servers, static AST analyzers, and PostgreSQL pgvector RAG stores.

Step 1

LLM Agent Client

Claude Desktop / Cursor / Web Agent

< 2msInspect →
Step 2

MCP Server Gateway

sparrow84001/mcp-seo v1.0.4

< 5msInspecting
Step 3

AST & Static Analyzer

Framework-Aware Parser

18ms – 40msInspect →
Step 4

Amazon Bedrock

Titan Text v2 Embeddings

80ms – 160msInspect →
Step 5

PostgreSQL pgvector

HNSW Cosine Vector Index

< 14msInspect →
Protocol

MCP Server Gatewaysparrow84001/mcp-seo v1.0.4

High-performance TypeScript server built on Bun runtime, exposing Zod-validated auditing and framework discovery tools.

Latency< 5ms
Scale / ThroughputSub-second Dispatch
Runtime: Bun v1.4
Schema Engine: Zod v3
Tool Registry: Dynamic multi-tool dispatch
Scientific Computing • AI Infrastructure • Open Source

Engineering Case Studies & Repositories

Verifiable research artifacts from Jefferson Lab & BNL alongside production AI architectures and open-source MCP tooling from @sparrow84001.

Particle Physics & Deep LearningSource

CaloGAN: Accelerating Detector Simulation 100x via Deep GANs

Brookhaven National Lab (BNL) & Jefferson Lab ATHENA Collaboration

Mentored by Dmitry Romanov (Jefferson Lab) within the Software & Computing Working Group.

The Engineering Challenge

Classical Geant4 Monte Carlo simulations model particle showers step-by-step, taking hours per event and creating massive computational bottlenecks for collider physics research.

System Architecture

Generative Adversarial Network (GAN) trained on Jefferson Lab calorimeter energy deposit showers, substituting classical ray-tracing with one-shot neural tensor generation.

Up to 100x acceleration in simulation throughput with preserved shower lateral profile fidelity.
PythonC++CaloGANGeant4ESCALATE FrameworkPyTorch/TF
AI Infrastructure & AST EngineSource

mcp-seo: Framework-Aware Model Context Protocol Server

Open-Source AI Tooling Engine on Bun Runtime

Authored and published by Sayanta Neogi as an open-source MCP server for AI coding agents.

The Engineering Challenge

AI development assistants lack AST-level static awareness of routing architectures, canonical URL consistency, and framework metadata across Astro, Next.js, and Laravel.

System Architecture

Bun-native Model Context Protocol server exposing Zod-validated audit tools, framework discovery analyzers, and non-destructive unified diff patch generators.

Sub-40ms static AST walks and automated verified code patch emission.
TypeScriptBun v1.4Model Context ProtocolZod v3AST Visitors
Multi-Agent Systems & Socratic AISource

vartualT: Socratic Multi-Agent Intelligent Tutoring Engine

Strict Pedagogical AI Guided by Conceptual Decomposition

Engineered as an autonomous educational platform that avoids direct answer hallucination.

The Engineering Challenge

Commercial LLMs frequently dump direct answers to homework problems, destroying critical thinking and failing to detect conceptual prerequisites.

System Architecture

Multi-agent supervisor pattern with dedicated hint-generation, misconception-diagnosis, and Socratic questioning agents orchestrating student guidance.

100% adherence to Socratic inquiry with zero raw solution leaks.
PythonLangChainMulti-Agent SupervisorsFastAPIPrompt Engineering
Enterprise AI & Vector Storage

Enterprise Bedrock + pgvector RAG Pipeline

Sub-Second Semantic Retrieval Over 500k+ Document Chunks

Architected and deployed at Web Mind Ltd with strict enterprise tenant isolation.

The Engineering Challenge

High-throughput unstructured knowledge retrieval across diverse enterprise documents while upholding zero-trust data confidentiality agreements.

System Architecture

Amazon Bedrock Titan Text v2 embeddings indexed via PostgreSQL pgvector with HNSW cosine distance (<=>) and VPC endpoint isolation.

<14ms database lookup latency across 500,000+ vector embeddings.
Amazon Bedrockpgvector (HNSW)PostgreSQLLangChainDockerAWS
Open-Source Repositories (github.com/sparrow84001)
View all 24 repos on GitHub
Engineering History

Professional Experience & Research

4+ years engineering scalable full-stack applications, production LLM systems with Amazon Bedrock & pgvector, and particle detector simulations at national laboratories.

Web Mind Ltd

Software Engineer (Full Stack + AI & DevOps)

2026 – Present
  • Architected enterprise Laravel, React, and Python systems with high throughput and strict data isolation.
  • Designed and deployed production RAG pipelines using Amazon Bedrock, pgvector, and LangChain with sub-second retrieval.
  • Built autonomous AI agents and document intelligence workflows for PDF, image, and knowledge-base extraction.
  • Optimized PostgreSQL and vector database indexes for high-concurrency AI search workloads.
  • Engineered scalable cloud architectures across AWS, Azure, and GCP with strict zero-trust confidentiality controls.
Amazon BedrockLangChainpgvectorReactLaravelPythonAWSDocker
UKtekIndia

Full Stack Web Developer, AI & DevOps Engineer

2021 – 2026
  • Engineered dynamic full-stack web applications utilizing Laravel, Flask, and Django with responsive React frontends.
  • Integrated GraphQL and RESTful APIs for high-throughput, low-latency client-server interactions.
  • Implemented generative AI features using LangChain and Hugging Face models for NLP and conversational intelligence.
  • Built intelligent agent systems and RAG architectures for enhanced search and knowledge retrieval.
  • Established automated CI/CD pipelines using GitHub Actions and GitLab CI/CD for containerized Kubernetes deployments.
  • Integrated Vite as frontend build tool for faster build times and developer velocity.
ReactLaravelDjangoFlaskKubernetesDockerGraphQLCI/CDLangChain
BNL + Jefferson Lab | ATHENA | CaloGAN

Software & Computing Working Group

2021 – 2022
  • Developed calorimeter simulation pipelines using Generative Adversarial Networks (GANs) for high-energy particle physics.
  • Collaborated with international scientific research teams under the supervision of Dmitry Romanov (Jefferson Lab).
  • Accelerated computational simulations using Python, C++, and deep generative models.
CaloGANGenerative Adversarial NetworksPythonC++Scientific Computing
BNL + Jefferson Lab | Electron-Ion Collider

Research Developer

2020 – 2021
  • Utilized the ESCALATE framework and Geant4 for particle physics detector simulations and analysis.
  • Developed high-performance simulation software using C++ and Python for particle collision modeling.
  • Published code and simulation artifacts under the mentorship of Dmitry Romanov.
Geant4ESCALATE FrameworkC++PythonDetector Modeling
Academic Foundations
2019 – 2021

Master of Science in Computer Science

Ramakrishna Mission Residential College, University of Calcutta

2015 – 2019

Bachelor of Science in Computer Science

Kanchrapara College, University of Kalyani

Technical Stack

Skills & Technologies

Full-stack web engineering, cloud-native containerized architectures, and specialized AI frameworks including Amazon Bedrock, pgvector, and Model Context Protocol.

AI Engineering & LLMs

Agent DevelopmentModel Context Protocol (MCP)RAG PipelinesAmazon BedrockLangChainHugging Facepgvector & Vector DBsPrompt EngineeringDocument IntelligenceTensorFlowGoogle ADK

Full Stack & Web

ReactAstroTypeScriptJavaScriptHTML5 & CSS3ViteTailwind CSSLaravel (PHP)Django & Flask (Python)ASP.NET (C#)GraphQL & REST APIs

Cloud & DevOps

Docker & ContainersKubernetesGitHub ActionsGitLab CI/CDAWS (Bedrock, S3, RDS, Lambda)AzureGCPIBM Cloud.NET Aspire

Databases & Storage

PostgreSQL & pgvectorMySQLMongoDBOracleMicrosoft SQL ServerVector Databases

Systems & Quantum

C / C++RustQ#QiskitCirqTensorFlow QuantumOpenMPCUDADPC++Geant4 SimulationReact Native
Get in Touch

Initiate a Conversation

Send an engineering inquiry, collaboration proposal, or project discussion. I review all inquiries directly and typically reply within 24 hours.

Direct Engineering Channels

Reach out directly for AI engineering, full-stack web platforms, or cloud DevOps roles.

Current Availability

Open for senior AI/ML engineering, scalable web architectures, and cloud consulting contracts worldwide.