Gurugram, India

Hi, I'm Om Raj.

I am a Backend Engineer

Software Development Engineer at Analytics Vidhya. I build multi-tenant Django backends, agentic RAG pipelines and event-driven data systems that keep running in production.

om@raj: ~/profile
0
REST APIs shipped in one multi-tenant backend
0
learners served across enterprise and consumer tiers
0
AI responses routed across 20+ intents
0
LLM tokens handled by the RAG pipeline
0
AI summaries generated by nightly pipelines
0
scheduled jobs automated with alerting

$ cat about.md

Portrait of Om Raj
Om Raj

I'm a backend engineer who likes systems that are boring in the best way: predictable, observable and easy to change. At Analytics Vidhya I work on Mentornaut, an AI learning platform, where I own the multi-tenant Django REST backend, the LLM and RAG layer, and the pipelines around them.

I started there as an SDE intern in January 2025 and moved to a full-time SDE-1 role in July 2025. Before that I earned my B.Tech in Electronics and Communication Engineering from NIT Patna (CGPA 8.04).

Outside work I practise competitive programming (CodeChef 4 Star, LeetCode Knight) and build side projects that scratch a real itch, like asking questions to a GitHub repository in plain English.

roleSoftware Development Engineer 1
companyAnalytics Vidhya
focusDjango, LLM / RAG, event pipelines
locationGurugram, India

$ git log --experience

Software Development Engineer 1

Analytics Vidhya / Gurugram, India

Jul 2025 - Present
  • Architected a multi-tenant Django REST Framework backend for Mentornaut, an AI learning platform with 45+ REST APIs serving 5,700+ learners across enterprise and consumer tiers; a Redis-cached routing middleware resolves each request's tier so one change ships everywhere.
  • Built a LangChain/LangGraph agentic RAG pipeline integrating 5 external AI agent services, routing 13,100+ responses across 20+ classified intents over 31.5M LLM tokens with OpenAI/Anthropic/Gemini fallbacks and Qdrant vector retrieval.
  • Shipped an AI assessment engine and nightly Celery summarization pipeline generating 905 MCQs and 5,800+ AI summaries; added stuck-job recovery, leaving 0 of 92 attempts stranded in grading.
  • Built a Kafka consumer (Avro + Schema Registry) ingesting course-enrollment events into PostgreSQL with Redis-buffered batching, and automated 25+ Celery Beat jobs with Sentry and MS Teams alerting.
Django RESTLangGraphRAGQdrantCeleryKafkaRedisPostgreSQLSentry

SDE Intern

Analytics Vidhya / Gurugram, India

Jan 2025 - Jun 2025
  • Migrated the Free Courses system from file-based storage to a PostgreSQL database, building a Django REST backend with JSON schema validation; 2 teams now manage 120+ courses dynamically without GitLab commits, cutting deployment time from hours to minutes.
  • Rebuilt Brahma, a user activity tracking system, from Django to FastAPI with token-based authentication, capturing 50,000+ monthly behavioral events with 95% accuracy using GeoIP2 and an AWS Kinesis-S3 pipeline for async event storage.
  • Eliminated cross-service data sync issues between NewCompetition and QuizMania with a PostgreSQL Foreign Data Wrapper for real-time user data access via the API Gateway, removing data duplication.
FastAPIDjango RESTPostgreSQL FDWAWS KinesisS3GeoIP2

$ ls ./skills

Python C++ Java SQL Django Django REST Framework FastAPI Celery SQLAlchemy GraphQL Django Channels PostgreSQL MySQL Redis Qdrant pgvector RAG Systems LangChain LangGraph OpenAI Anthropic Claude Google Gemini Langfuse AWS (S3, Kinesis) Docker Kafka Git GitLab Jenkins CI/CD Sentry Postman Grafana Metabase JWT Gunicorn Whitenoise GeoIP2 Pydantic

$ ls ./projects

SealGrade

Tamper-resistant code grading, an exploit corpus and an auditor

Oct 2026
agentartifact firewallexecutionjudgesigned verdict
  • Built an open-source grading runner in Python and Docker: three isolated containers (non-root, no network, read-only filesystem, dropped capabilities), a single-file artifact firewall, a type-strict judge that never runs candidate code, and HMAC-signed verdicts.
  • Wrote a corpus of 30 attacks against grading harnesses and measured it: a naive harness fell to 21 of 21 exploits, the strict tier to 0. Every claim is asserted in CI (226 unit tests, Hypothesis fuzzing, Docker integration tests).
  • Shipped a static and mutation-testing auditor with 26 rules, SARIF output for GitHub code scanning, a reusable GitHub Action, a MkDocs site and a tagged release.
PythonDockerPydanticHypothesisGitHub ActionsMkDocs
View on GitHub Read the docs

Code-Query

Chat with any GitHub repository

Jun 2026
repochunksembeddingspgvectoranswer
  • Engineered a Retrieval-Augmented Generation application enabling developers to query GitHub repositories in natural language, chunking source files into 768-dimension vector embeddings for semantic search.
  • Architected an asynchronous ingestion pipeline using FastAPI and SQLModel generating embeddings via Gemini text-embedding-004, stored in PostgreSQL with pgvector for cosine similarity search; Dockerized with JWT auth.
PythonFastAPILangChainPostgreSQLGemini AIDocker
View on GitHub

HireHub

Role-based job platform

Jan 2026
AdminStaffCandidateAppliedBookmarked
  • Built a job platform with Role-Based Access Control (Admin, Staff, Candidate) across 10+ Django REST Framework APIs and a React.js frontend (React Router v7, Zustand, Tailwind CSS).
  • Architected a normalized PostgreSQL schema with User-Job mapping tables to track application statuses (Applied, Bookmarked, Clicked), implementing django-filter for search across title, company, location and tags.
Django RESTReact.jsPostgreSQLZustandTailwind
View on GitHub
$ ls ./earlier-projects (9)

$ cat contests.log

  • #108 of 26,446 in CodeChef Starters 115 proof
  • #1391 of 22,688 in LeetCode Weekly Contest 383 proof
  • #3,224 of 30,822 in Codeforces Hello 2024 proof

$ cat education.txt

National Institute of Technology Patna

Bachelor of Technology, Electronics and Communication Engineering

May 2025

CGPA 8.04

DAV Public School, Gaya

CBSE, Intermediate (12th) and Matriculation (10th)

2018 - 2020

12th 94.0% / 10th 95.4%

$ mail --to om

compose

$ mail -s "hello" omraj010@gmail.com

# Open to conversations about backend, LLM systems and interesting problems.