Full-Stack & AI Engineer · India

Prakul
Parashar

Building scalable backends, agentic AI systems, and useful AI-powered products.

Node.js · Express · PostgreSQL / MySQL · LangChain · LangGraph · LLMs · RAG ·LLM Evaluation · multi-agent systems · MCP.

About

Robust server-side architecture, meeting stateful AI.

I work at the intersection of backend engineering and applied AI — designing secure authentication, scalable backend logic and data models on one side, and multi-agent orchestration, tool-using workflows and retrieval systems on the other.

Most of my work involves stateful AI: graphs of agents that call tools, validate their own outputs,Undergo LLM Evaluation recover from bad retrievals, and keep behaving predictably once real users and real data arrive.

"Build useful AI-powered software — not simple AI demos."
Focus
Backends · Agentic AI · Automation
Approach
Schema-first, production-oriented
Currently
B.Tech IT · building AI systems

Education & Experience

A short, deliberate track record.

Education

  1. 2024 — 2028

    B.Tech, Information Technology

    Maharaja Agrasen Institute of Technology (GGSIPU)

    Core computer science, systems and applied AI engineering.

  2. Completed 2024

    Class 12, Higher Secondary Education

    Ch. Baldev Singh Model School (Sr. Sec.)

    Science stream with a focus on mathematics and computing.

Experience

  1. Jul 1 — Jul 31, 2026

    Software Engineering Intern — NCR Atleos

    Applied AI · Hiring workflows

    Developed an AI Resume Screener to streamline candidate evaluation — parsing resumes into structured data, matching them against job descriptions and surfacing ranked shortlists for recruiters.

Skills

The stack I reach for.

Grouped by the part of the system it belongs to.

01

Backend & APIs

  • Node.js
  • Express.js
  • Python
  • PostgreSQL
  • MySQL
  • Redis
  • REST APIs

02

AI Engineering

  • LLMs
  • RAG
  • Embeddings
  • LangChain
  • LangGraph
  • LLM evaluation & guardrails
  • Multi-agent workflows
  • Tool-calling systems

03

Data & Reliability

  • Pydantic
  • TypeDict
  • Schema validation
  • Structured outputs
  • Stateful workflows

04

Integration & Infrastructure

  • JWT
  • OAuth
  • Redis clustering
  • MCP servers/clients
  • Web scraping
  • Async API integration

LeetCode

Live problem-solving metrics.

Pulled directly from my LeetCode profile.

Fetching live stats…

Services

How I can help.

  • 01

    AI Engineering

    LLM applications, RAG pipelines, embeddings, agentic workflows, offline evaluation pipelines and multi-agent systems designed for real usage.

  • 02

    Backend Development

    Scalable APIs, database-backed systems, authentication and stateful application logic.

  • 03

    Intelligent Automation & Data Integration

    AI tools connected to APIs, databases, search, scraping and external services through MCP.

  • 04

    Full-Stack Product Development

    End-to-end, production-oriented AI-powered applications — from schema to interface.

Selected Projects

Systems, not demos.

AI Resume Screener

01
Problem
Recruiters spend hours reading resumes that do not match the role.
Solution
Extracts structured information from resumes, compares candidates against job descriptions, ranks them and answers recruiter-focused questions over the applicant pool.
  • Python
  • LLMs
  • Structured outputs
  • Pydantic
  • RAG

Multi-Agent Financial AI System

02
Problem
Investment research requires several distinct analytical competencies.
Solution
Specialised agents handle market analysis, portfolio allocation, security selection and report generation, coordinated as a stateful graph.
  • LangGraph
  • LangChain
  • Tool calling
  • Python

Corrective RAG System

03
Problem
Naive retrieval confidently answers from irrelevant documents.
Solution
Grades retrieved documents, rewrites weak queries, falls back to web search when retrieval fails, refines the final context before generation & and undergo offline LLM evaluation.
  • LangGraph
  • Embeddings
  • Vector search
  • Web search
  • DeepEval

Web Scraping MCP Server

04
Problem
Scraping tools are glued to agents with bespoke, brittle code.
Solution
A custom Model Context Protocol server that exposes scraping capabilities as standardised tools for any MCP-aware LLM client.
  • MCP
  • FastMCP
  • Python
  • Async I/O
  • Scraping

Delhi Geospatial Accessibility Analysis

05
Problem
District-level infrastructure gaps are invisible in aggregate statistics.
Solution
Evaluates districts against population-based facility standards using OpenStreetMap data and walking-network analysis to quantify coverage and gaps.
  • OSMnx
  • OpenStreetMap
  • Python
  • Geospatial analysis

E-Commerce Marketplace

06
Problem
Monolithic storefronts lack real-time data sync and struggle with scalable inventory and order handling.
Solution
A full-stack marketplace featuring dynamic product cataloging, secure user authentication, and relational order management built on a decoupled PERN stack.
  • Node
  • Express
  • REST APIs
  • PostgresSQL
  • Redis
  • JWT
  • OAuth

Contact

Let's build something that ships.

Open to collaboration on AI systems, scalable backends, intelligent automation and full-stack products.