Available for Junior AI & Full-Stack Roles

Engineering Intelligent AI Agents & Robust Systems

I specialize in designing and building production-ready architectures: from enterprise RAG search engines and secure Docker-sandboxed code execution environments, to high-performance real-time financial intelligence platforms.

Try Interactive Simulator View My Projects

3+ Core AI Architectures

100% Sandboxed Execution

Real-Time Data Integrations

agent_system_orchestrator.py

# Orchestrating multi-agent tool pipeline import docker, langchain, openai class AgentOrchestrator: def __init__(self): self.sandbox = docker.from_env() self.agent = langchain.agents.initialize() def execute_sandboxed_code(self, script_code): container = self.sandbox.containers.run( image="python:3.10-slim", command="python -c " + script_code, network_mode="none", # Secure air-gap mem_limit="128m", detach=True ) return container.wait(timeout=10)

Technical Expertise

My engineering toolkit is structured to span the entire lifecycle of modern intelligent software applications, prioritizing security, analytics, and speed.

AI & LLM Architecture

Retrieval Augmented Generation (RAG)
Vector Stores & Indexing (FAISS)
LLM Orchestration (LangChain)
API Integration (OpenAI, Gemini SDK)
SentenceTransformers & Embeddings
Agentic Workflow Design

Full-Stack Engineering

React 18 & 19 (Hooks, Suspense)
Node.js & Express servers
FastAPI (Python web framework)
Vanilla HTML5, CSS3, ES6+ JS
Vite, bundlers & builds
Tailwind CSS v4 & custom HSL setups

Data & Databases

Pandas & NumPy data pipelines
MongoDB document database
SQLModel & SQLite relational databases
Vector Embeddings storage
OpenPyXL spreadsheet analytics
JSON schema validation

DevOps & Environments

Docker & container sandboxing
Docker Compose orchestration
Secure sandboxed environments
Linux server environments
Git version control & workflows
REST API security & Auth

Core Systems Portfolio

These are three highly complex applications I engineered, demonstrating advanced integration of AI agents, RAG engines, sandboxing, and full-stack design.

📈 e:/PROJECT/marketminds

MarketMind — Real-Time Stock Analytics & AI Trading Coach

A sophisticated financial analytics dashboard and journal. Integrates the Google GenAI SDK to review trade histories, behavioral logs, and emotional states (such as FOMO or revenge trading), delivering multi-dimensional institutional coaching reports, letter-grade metrics, and time-of-day analytics.

Gemini 3.5 Flash: Behavioral trade diagnostics.
Motion Engine: Fluid framer-motion micro-interactions.
Social Feed & Leaderboard: Live consistency rankings and user posts.
State Sync: High-integrity MongoDB state synchronization.

Simulate Gemini Analysis

📊 e:/PROJECT/StatBot

StatBot Pro — Secure Sandboxed AI Spreadsheet Analyst

An advanced CSV/Excel data analysis agent. It takes analytical commands, uses Langchain + OpenAI to write custom Python data manipulation scripts, and executes them inside an isolated, air-gapped Docker container for secure, sandboxed data processing.

Docker SDK: Spins up transient air-gapped agent containers.
Langchain Code Agent: Dynamically generates Pandas/NumPy scripts.
Structured SQLite: Persists data using SQLModel schemas.
Reactive Frontend: Interactive tabular and graphical output renderers.

Simulate Docker Sandbox

📄 e:/PROJECT/DocuMind

DocuMind — Enterprise-Grade RAG Knowledge Assistant

A fast, localized Retrieval-Augmented Generation assistant. Users upload PDF documents, which are parsed with PyPDF, segmented, and embedded using SentenceTransformers into a local high-performance FAISS vector store to enable precise grounded Q&A via Groq LLMs.

FAISS Vector Store: Ultra-low latency similarity searches.
SentenceTransformers: Lightweight embeddings generated locally.
FastAPI Stream: Efficient routing structure for uploads and queries.
Air-Gapped Ready: Highly portable structure adaptable for local LLMs.

Simulate RAG Retrieval

Project Architecture & Interactive Simulator

Interact with simulated versions of my applications below to see exactly how their internal architectures operate in real time.

MarketMind Simulation

Simulates executing a Gemini AI diagnostic on a trader's transaction log to recognize psychological and temporal habits.

Select Trader Journal Preset:

Trades Count 8

Win Rate 37.5%

Net Profit -$420.00

StatBot Pro Simulation

Simulates Langchain generating Python code and executing it within a secure, isolated Docker container to analyze an uploaded CSV.

Select CSV File:

What should the AI check?

DocuMind Simulation

Simulates similarity search using local FAISS vector store indices and generating grounded answers via Groq API.

Document Index Grounding:

Select or Type Query:

Console: Idle

Ready. Select a simulator preset and click run.

Engineering Journey

A progression through building full-stack web products to managing complex localized intelligence platforms and Docker containers.

Late 2025 - Present

Localized AI & RAG Orchestration

Engineered FAISS-grounded vector search engines and sentence embedding structures locally, cutting cloud costs by 80% and keeping enterprise data private. Focused on optimization of chunk size strategies, indexing, and FastAPI high-concurrency routes.

Mid 2025

Secure Agentic Environments

Designed and built StatBot Pro, resolving the safety concerns of executing LLM-generated code by creating transient, air-gapped Docker execution chambers on-demand. Managed secure file-mounting systems and resource throttle limits.

Early 2025

Financial Intelligence Dashboards

Synthesized Express backend states, Vite frontend speeds, and MongoDB data pipelines to construct MarketMind. Integrated Gemini SDK to diagnose trading psychology, building custom consistency calculations to score performance.

2024 & Prior

Full-Stack System foundations

Developed core competencies in database schemas, REST APIs, responsive vanilla CSS engineering, and stateful React applications. Built web services, integrated authentication schemas, and optimized build setups.

Let's Build Something Impactful

I thrive on solving complex architectural challenges: whether that is optimizing retrieval performance, building air-gapped agentic workflows, or engineering real-time data sync services.

Direct Communication nikhildubale2405@gmail.com

Open Source Work github.com/nikhildubale