IHAR IVANIUK

Discuss a project

Public Mindshine / TIVA website, shown for company context. The internal product described in this case is not shown.

Public Mindshine / TIVA website, shown for company context. The internal product described in this case is not shown.

Mindshine

TIVA

Full-stack Developer · AI Platform

TIVA — Mindshine’s voice and text retrieval product — reaches data regardless of source so teams can query knowledge without a siloed search screen.

01Problem

Voice and text retrieval had to span heterogeneous sources and stay inside existing workflows, including RPC paths that had to remain stable when load increased.

02Selected contributions

Built client and server application layers for TIVA: React on the client, Node.js services over RPC, retrieval data in PostgreSQL and MongoDB. Profiled and debugged RPC paths under load. Assisted technical direction and coding practice — did not solely own the platform architecture.

  • Client and server application layers for voice and text retrieval
  • RPC profiling and debugging under load
  • Persistence against PostgreSQL and MongoDB
  • Technical direction and coding practice alongside the team

03System context

React clients call application services over RPC. Retrieval data lives in PostgreSQL and MongoDB. Cube and Ansible sit in the analytics/delivery stack around the product; named Cube models are not published.

  • React clientsMy scopeCalls the APINode.js services over RPCMy scope
  • Node.js services over RPCMy scopePersists dataPostgreSQL · MongoDB

My scope in the system: React clients, Node.js services over RPC

Delivery

  • AnsibleDelivery pipeline

Separate from the runtime flow. These are not connected to each other.

Implementation tools

Cube (analytics tooling, not a runtime hop). Not runtime services.

04Scale & constraints

  • Heterogeneous sources feeding a knowledge base
  • Voice and text retrieval in one product
  • RPC stability under load
  • Client and server in the same engagement; platform architecture was not solely owned

05Outcome

Client and server layers for the retrieval product shipped, and RPC paths were profiled under load. Latency numbers and source-coverage figures are not published.

  • Production client and server layers for the retrieval product
  • RPC paths profiled under load

06Stack

  • TypeScript
  • React
  • RxJS
  • Ramda
  • Node.js
  • RPC
  • PostgreSQL
  • MongoDB
  • Cube
  • Ansible

No verified public metrics are published for this engagement.

Categories: AI, SaaS

Year: 2024

Open live project ↗

Related services