Raj Goswami

About

I think in systems.
I build in layers.

I started in backend logic — debugging pipelines, tracing wrong outputs back to wrong assumptions. Somewhere in that process I realized the most interesting layer wasn't the code. It was the instruction: what you tell a system to do, and how precisely you say it.

Today I work at the intersection of full-stack development and prompt engineering. I design structured prompts that reduce model ambiguity, build scalable MERN stack applications that integrate AI outputs into real workflows, and obsess over edge cases most people skip.

My thesis is simple: intelligent systems don't fail at the model level — they fail at the interface between human intent and machine interpretation. That's the gap I close.

2+ Years building

3 Internships

∞ Edge cases studied

Capabilities

What I work with

AI & Prompt Engineering

Development

Data & Systems

Tools

Selected Work

Projects

Recent Work React · Node.js · MongoDB · OpenAI API

AI Interview Preparation Platform

Resume Analysis

Built a MERN-based platform for resume analysis and interview preparation, implementing advanced resume parsing to extract structured insights and provide optimization suggestions.

AI Integration

Integrated the Google Gemini API to dynamically generate personalized mock questions and immediate feedback tailored to the user's specific skills and experience.

Jan – Apr 2026 React · Node.js · Express · MongoDB

Water & Electricity Tracking App

Final Year Project

Dashboard & Insights

Engineered dynamic dashboards using React.js with interactive data visualizations for monitoring real-time water and electricity usage, enhancing user decision-making.

Backend Architecture

Designed and implemented high-performance REST APIs using Node.js and Express.js for scalable data handling.

Anomaly Detection

Integrated intelligent anomaly detection mechanisms to identify irregular consumption patterns and alert users to potential faults or leaks.

Sep – Nov 2025 MERN Stack

E-Commerce Web Application

Core Platform

Developed a highly scalable full-stack e-commerce platform featuring a robust cart system and comprehensive product management.

Security

Implemented secure authentication flows using JWT-based login, ensuring strict role-based access for administrators and customers.

Optimization

Built highly efficient REST APIs and optimized database queries to ensure smooth performance during heavy concurrent usage.

Prompt Engineering

How I work with LLMs

Three case studies showing structured thinking — not just prompt guessing.

01

Structured Prompt Templates

Instead of free-form prompts, I design templates with explicit role definitions, output format constraints, and failure conditions baked in.

prompt-template.txt

ROLE: You are a data analysis assistant.
CONTEXT: {user_data_summary}
TASK: Identify top 3 anomalies in the dataset.
CONSTRAINTS:
  - Output ONLY a JSON array
  - Each item: { "anomaly": str, "severity": 1-5 }
  - Do NOT explain reasoning unless asked
FALLBACK: If data is insufficient, return []

Result Consistent, parseable outputs with ~90% less post-processing needed.

02

Output Improvement Iterations

I analyze model outputs systematically — not by eye — identifying where ambiguity in the prompt leaked into the answer.

v1

Prompt: "Summarize this document."

Output: 400-word verbose recap with no structure.

↓ identify: length & format ambiguity

v2

Prompt: "Summarize in 3 bullet points. Each ≤ 20 words. Technical audience."

Output: 3 precise, scannable bullets. Zero fluff.

03

Edge-Case Handling

Production AI systems break on inputs the designer didn't anticipate. I design for those inputs explicitly.

Empty Input → Return structured "no data" response, not an error

Contradictory Instructions → Prompt includes priority hierarchy: safety > accuracy > style

Language Mismatch → Explicit locale constraint in system message

Out-of-Scope Query → Graceful declination with suggested redirect

Experience

Where I've worked

Full Stack Developer Intern

InternPe

Aug – Sep 2024

Developed responsive single-page applications using React.js, improving UI performance and user experience.
Built RESTful APIs using Node.js and Express.js for scalable backend services.
Implemented secure authentication using JWT.
Worked with API integrations and optimized application performance in an Agile environment.

Software Developer Intern

BharatIntern

Mar – Jun 2024

Developed backend services using Node.js and integrated MongoDB databases.
Improved data processing and validation logic, ensuring data consistency.
Debugged and tested applications for substantial performance improvements.

Position of Responsibility

Udessya Club (NGO), KIET

Mar 2023 – 2024

Participated in community service and managed event coordination.
Developed leadership, communication, and strong teamwork skills.

B.Tech (Computer Science)

KIET Group of Institutions, Ghaziabad

Aug 2022 – Jun 2026

Class X & XII (Science)

Gramodaya International College, Jhansi

2018 – 2021

Certifications

✓ Full Stack MERN Development — Sheryians Coding School

✓ Certification in Ubuntu Linux: Operating System Basics

Get in touch

Let's build something
worth building.

Whether it's a prompt engineering challenge, a full-stack integration, or something that doesn't fit either box — reach out.

📞 +91 7054309709 ✉ goswamiraj2468@gmail.com in linkedin/raj-goswami003 { } github/RajGoswamics

I think in systems.I build in layers.