Designing production systems where scale, latency, and failure are first-class concerns.

name: Vansh Taneja
role: Software Engineer (AI + Distributed Systems)
education: Computer Science (AI Option), University of Alberta
operating_mode: production-first
focus:
  - Event-driven architecture
  - Real-time decision systems
  - AI under latency constraints

I don’t just build demos.
I design systems with explicit failure modes.

Aether — Real-Time Distributed Logistics Engine

Dispatch is not CRUD.
It is a streaming, stateful decision problem under strict latency constraints.

• Kafka-backed event backbone
• gRPC microservices with strict service contracts
• Redis for real-time fleet and order state
• PostgreSQL for durability and analytical workloads
• AI models embedded directly in the decision loop
• Operator command-center interface for live overrides

🔗 https://github.com/vanshtaneja23/Aether

Event-Driven Video Streaming Platform

• Asynchronous transcoding pipelines
• Queue-isolated workloads
• Object-storage-backed scalability
• Container-first deployment model

🔗 https://github.com/vanshtaneja23/micro-netflix

Federated Social Networking (ActivityPub-style)

Multi-node, REST-first. Entries, likes, and comments flow node-to-node via inbox APIs.

• RESTful API with full spec compliance; pagination, auth, visibility (public/unlisted/friends)
• Node-to-node federation: follow requests, entries, likes, comments pushed to remote inboxes
• PostgreSQL-backed; single-server-per-node deployment model
• Every endpoint tested and documented; interop with multiple teams' nodes

Software Quality • Java REST Service • IoT Platform

Evaluating and extending an existing Java service layer. Quality and testability over new features.

• Java, Dropwizard/Jersey — RESTful state and update APIs for smart-home nodes
• Integration with IoT controller and historian; state consistency and failure recovery
• Focus: testing, requirements validation, and improving a legacy codebase

• LiteLLM — fixed a Gemini bug causing HTTP 500s, independently verified by an affected user (PR #32545, #31505)
• Neural Playground — MLP + backprop from scratch in pure JS, trains live in-browser (Live Demo · repo)
• LLM Orchestrator — self-hosted LLM gateway, SQLite request logging, tested failure paths (repo)

I take ownership of systems end-to-end.
I design for scale, reliability, and real-world constraints.
I build software that combines distributed systems with AI and machine learning.

Roles:
Software Engineering Intern • Backend / Distributed Systems • AI / ML Engineer

Contact: vanshtaneja23@gmail.com

Interested in building scalable infrastructure, intelligent systems, and production-ready AI.