Pan Science Innovations(PSI) | Full Stack Developer(FSD) Intern | On-Campus

Status: Final year B.Tech CSE student at IIIT Bhagalpur Position: Full Stack Developer (FSD) Intern Result: Rejected via official college TPO email after clearing all engineering rounds.

Round 1: Full-Stack Take-Home Assignment & Architecture Video

The interview process began with a rigid full-stack assignment request.

• The Task: Build an end-to-end task management platform from scratch featuring strict Role-Based Access Control (RBAC).
• My Implementation: I built the application, handled the frontend styling, coded the core backend logic, deployed the entire architecture live on AWS infrastructure, and configured automated continuous integration/continuous deployment (CI/CD) pipelines using GitHub Actions.
• Submission Requirement: Along with the code repository, I had to submit an exhaustive 13-minute video documentation walking through the operational codebase, database architecture, security middleware implementation, and active deployment configurations.

Round 2: Technical Interview with Engineering Managing Director (MD)

Out of the pool of applicants, I was the only candidate selected to advance to the final technical round with the Engineering MD. Although I applied for a traditional Full Stack Developer role, the interview leaned significantly into infrastructure hardware mechanics and AI systems.

• Hardware & Systems Architecture:
  • Question: Why do modern AI/deep learning models run almost exclusively on GPUs rather than traditional CPUs?
  • Answer: I explained the architectural divergence between serial processing execution chains in CPUs (designed for deep single-task logic processing cores) versus highly parallel processing distribution structures across thousands of smaller, efficient cores inside modern GPUs, making them optimal for large-scale matrix multiplications.
• AI Pipelines:
  • A comprehensive mapping explanation of data flows within typical Retrieval-Augmented Generation (RAG) platforms.
• System Design Scalability Scenario (CCTV Edge Case):
  • Scenario: Suppose you have a distributed cluster of 10,000 CCTV cameras across multiple physical environments generating continuous video streams. You are provided with their respective geo-coordinates (latitude/longitude), and many video captures are overlapping. How would you design a software architecture to efficiently parse these streams to track down a single human target wearing a red hat?
  • My Approach: Initially, I pitched an abstract chunking data processing strategy passing data contexts, which was immediately corrected by the MD. I pivoted, drawing directly from my experience at Smart India Hackathon (SIH) involving density-based intelligent traffic frameworks. I structured an architecture utilizing edge computing processing nodes running a computer vision pipeline with custom-trained YOLOv8 object detection algorithms to isolate target features and extract automated localized telemetry updates.
• Project Review: The MD confirmed he had watched my complete 13-minute architecture assignment video breakdown from start to finish. He gave incredibly positive feedback regarding my communication clarity and engineering documentation style. I spent an additional 10 minutes walking systematically through my app's boilerplate structure, operational middleware layers, routing mechanisms, and AWS deployment configurations without a single cross-question objection from his end.

The Outcome

The interview concluded on a highly positive note, with the MD explicitly stating that my file was being cleared to HR for final onboarding instructions and location tracking management.

Unfortunately, despite being the only candidate to clear the engineering gauntlet to the finish line, an official corporate rejection notice was delivered to my college campus placement cell a few days later.

My Takeaway & Key Learnings

1. Be prepared for domain crossovers: Even if the job description says "Full Stack Developer", don't assume you won't be tested on AI system design, hardware constraints, or algorithmic pipelines in today's market.
2. Video execution matters: Spending extra effort producing clean, professional architecture walkthrough videos for assignments differentiates you instantly to engineering directors who value structural communication.
3. Nothing is finalized until the contract is signed: Verbal confirmations during final rounds regarding onboarding status can still fall through due to unexpected corporate resource allocations, budget resets, or localized hiring freezes. Treat every outcome as a springboard to jump right back into active data structures, system design, and project preparation frameworks.