Problem Statements

Participants may choose one of the following engineering challenges.

AI Engineer

Problem Statement 1: Self-Correcting RAG Pipeline
Build a Retrieval-Augmented Generation (RAG) system capable of detecting insufficient or contradictory retrieved context and intelligently re-querying, asking clarification questions, or returning low-confidence responses instead of hallucinating. Demonstrate measurable improvements in hallucination rates using an evaluation harness.

Problem Statement 2: Constrained Structured Extraction at Scale
Develop an AI-powered structured extraction system capable of converting noisy contracts, chat logs, or support tickets into strict nested JSON while maintaining over 90% schema-valid output without relying on hardcoded regular expressions.

Voice AI Engineer

Problem Statement 3: Interruption-Aware Voice Agent
Build a real-time conversational voice agent capable of naturally handling interruptions, updating context dynamically, and continuing conversations without breaking the user experience.

Problem Statement 4: Low-Latency Multilingual Voice Pipeline
Develop a multilingual voice-to-voice pipeline that supports seamless language switching during conversations while maintaining production-grade latency and delivering accurate real-time interactions.

Backend Developer

Problem Statement 5: Rate-Limited Multi-Tenant API Gateway
Build a scalable backend gateway capable of routing requests across multiple AI providers while supporting tenant-based rate limiting, automatic failover, and intelligent request queueing under high load.

Problem Statement 6: Event-Driven Pipeline with Guaranteed Delivery
Develop a distributed asynchronous job processing system with retry mechanisms, idempotency guarantees, dead-letter queues, and observability while ensuring zero data loss during worker failures.

Open Innovation

Problem Statement 7: Open Innovation
Have an idea beyond the predefined challenges?

 

Participants may propose their own engineering problem, provided it clearly belongs to one of the three domains — AI Engineer, Voice AI Engineer, or Backend Developer — and matches or exceeds the technical complexity of the official problem statements. Open Innovation submissions will be judged using the same evaluation criteria as every other project.

Judging Criteria

Projects will be evaluated on:

  • Engineering Design & Architecture
  • Functionality & Correctness
  • Technical Implementation
  • Code Quality
  • Performance & Scalability
  • Robustness & Reliability
  • Documentation
  • Live Demonstration
  • Overall Problem Impact
Organized By

OneInbox
OneInbox is an AI-native customer experience platform where AI agents and human teams work together across Voice, Chat, Email, and SMS. Trusted by 3,000+ users across 30+ countries and backed by Techstars and Upekkha, OneInbox is building the future of AI-powered customer communication.

Hosted By

AI House
AI House is the community layer powering this ecosystem. Built for founders, researchers, engineers, students, and AI enthusiasts, AI House has grown into a thriving network of 15,000+ builders, creating opportunities to learn, collaborate, and build the next generation of AI products.

Together, OneInbox and AI House are creating an ecosystem where ambitious builders transform ideas into production-ready products and careers.

Stages and Timelines

Round 1: Online Architecture Challenge

04 Jul 26, 09:00 AM IST → 22 Jul 26, 09:00 PM IST
Mode: Online

Participants must choose one problem statement from their selected domain (AI Engineer, Voice AI Engineer, Backend Developer, or Open Innovation) and submit a Product Requirements Document (PRD) along with a System Architecture Diagram using HiDevs Architecture Copilot.

Round 1 evaluates participants on problem understanding, product thinking, architecture quality, scalability, engineering decisions, technical feasibility, and documentation. No working implementation is required during this round.

Submission Requirements:

 

  • Product Requirements Document (PRD)
  • System Architecture Diagram
  • Selected Problem Statement
  • Solution Overview

    Duration: 4 July 2026 (9:00 AM IST) – 20 July 2026 (9:00 PM IST)
    Result: 22 July 2026 (9:00 PM IST)

    Live

Round 2: Offline Hiring Hackathon

  • 25 Jul 26, 10:00 AM IST → 25 Jul 26, 08:00 PM IST
    Mode: Offline
    Venue: OJONE, Bengaluru

    Shortlisted participants will build a complete working solution for their selected problem statement during the offline hackathon. Participants will demonstrate their solution, explain their architectural decisions, and participate in technical discussions with the OneInbox Engineering Team. Final internship selections will be made based on the overall evaluation.

    Submission Requirements:

    • GitHub Repository
    • Working Application / Deployable Demo
    • Project Description
    • Live Demonstration

     

    Duration: 25 July 2026 | 10:00 AM – 8:00 PM IST