Hi this Ayaan. i build intelligent backend systems

github

Ayaan Khan  • Peerlist

Built the 0-1 backend of the behavioural mock interview under my ownership which can take complete 45-60 mins of interview of candidate for a particular company and it's role. Implemented proctoring in interview to prevent any illegal activity while interview. Completely speech to speech conversation

Behavioural Mock Interview

Made a software which can help controlling the car racing game using hand gestures using camera.

GestureControlledGame

Problem: Developers lose valuable time managing repetitive tasks like code reviews, communication, and repository updates. • Solution: Built an AI agent that autonomously performs SDE-1 level tasks—GitHub operations (commit, PR, review), Slack/email context handling, and real-time voice interaction—through natural language over phone call. • Impact: Created a full-stack autonomous engineer that can interpret messages, perform actions, call the user for clarification, and self-pick tasks, significantly reducing manual overhead. • Enhancement: Developed IRIS CLI — a terminal-based AI assistant that leverages NLP to perform local actions like executing shell commands, editing and creating files, and automating workflows, all through natural language instructions.

IRIS – Autonomous AI Software Engineer

iris cli tool - in terminal assistant. pip install iris-ai-cli

Iris CLI

Built the 0-1 backend of the language mock interview under my ownership which can take complete 45-60 mins of interview of candidate for a particular language like (Java,Python,etc).Included 1 on 1 conversation plus code writing for coding questions. Implemented proctoring in interview to prevent any illegal activity while interview. Completely speech to speech conversation

Language Mock Interview

Made a personal ai assistant using the python and Openai's API key which can assist the user in many ways.

PersonalAssitant

<p>Made an Image Search Engine to get the images from local storage by just naturally describing the scene in images.</p>

Photosage

<p>Problem: No safe, low-privilege platform exists for practicing real Linux commands without risking host damage or privilege escalation. • Solution: Built an unprivileged containerized terminal sandbox using bubblewrap (10MB sandboxes, 3–5× lower RAM vs Docker), tmux (99.2% session stability), and cgroups for CPU/RAM throttling + runaway process protection (&lt;2s kill). • Engineering Highlights: * Cut sandbox startup to &lt;2s and enabled snapshot/restore for instant puzzle retries. * Added a PostgreSQL-backed ledger with 100% command traceability and events replay option. * Enforced security with 300+ blocked unsafe commands and namespace-based isolation. * Automated GC reduced orphan sessions by 95%. • Impact: Scaled to supports concurrent isolated sandboxes on a 1 GB VM, powering gamified Linux challenges, leaderboards, and reproducible puzzle environments.</p>

Shell Breaker — Interactive Linux Problem-Sol

Made a Tensorflow model which can convert the sign language to English in real time. Right now it is a prototype and it can convert only few sentences but in later updates it can perform more.

SignLangToEnglish

Eyes to Machine. Using opencv and Google's(Gemini) i managed to merge the power of AI using Gemini which allowed the machine to observe the environment around it.