Tekion | Staff Software Engineer | Interview Experience (AI Role)
Summary
I interviewed for a Staff Software Engineer (AI) role at Tekion, cleared three rounds covering DSA, low‑level design, and high‑level design, and received a positive outcome.
Full Experience
Tekion Staff Software Engineer | Interview Experience
Company: Tekion
Role: Staff Software Engineer (AI Role)
I recently interviewed for a Staff Software Engineer position at Tekion.
Round 1 (DSA)
Given multiple time intervals and the number of resources used during each interval, find the maximum total resources needed at any point in time.
Intervals:
[1, 4] → 3 resources
[3, 5] → 2 resources
[4, 6] → 4 resources
Output: 6
Solved it, and asked few questions related LLD and multi threading which was easy to explain
Round 2 (LLD)
Interviewer started asking HLD.
Design a ChatGPT‑style platform where users have conversations with an AI assistant. Users send messages, the AI's response streams back token by token, conversations are saved, and users can search across their chat history. The system serves many organizations (tenants), each with isolated data. Users can edit messages as well.
(Edit the previous posted question )
SS, represents my solution somewhat, It went well
Round 3 (HLD)
Design a platform that runs AI agents for other applications. An agent can call tools, run long (mins–hours), and may need human approval mid‑run. Multiple agent steps can be chained into a workflow (branches, loops, fan‑out). The platform must run these in isolated sandboxes, shared across tenants over a limited compute pool, and let callers either fire‑and‑forget or watch a run live step‑by‑step — without losing any data if a component crashes mid‑run.
This also went fine
They also asked a question about building an EVAL pipeline for a RAG system.
Interview Questions (4)
Maximum Resource Usage Over Intervals
Given multiple time intervals and the number of resources used during each interval, find the maximum total resources needed at any point in time.
Intervals:
- [1, 4] → 3 resources
- [3, 5] → 2 resources
- [4, 6] → 4 resources
Expected output: 6
ChatGPT‑style Conversational Platform (LLD)
Design a ChatGPT‑style platform where users have conversations with an AI assistant. Users send messages, the AI's response streams back token by token, conversations are saved, and users can search across their chat history. The system serves many organizations (tenants), each with isolated data. Users can edit messages as well.
AI Agent Execution Platform (HLD)
Design a platform that runs AI agents for other applications. An agent can call tools, run long (minutes to hours), and may need human approval mid‑run. Multiple agent steps can be chained into a workflow (branches, loops, fan‑out). The platform must run these in isolated sandboxes, shared across tenants over a limited compute pool, and let callers either fire‑and‑forget or watch a run live step‑by‑step — without losing any data if a component crashes mid‑run.
EVAL Pipeline for RAG System
Explain how you would build an evaluation (EVAL) pipeline for a Retrieval‑Augmented Generation (RAG) system.