Creative toolsneed realengineering.

I build the machinery creative tools run on: React and TypeScript frontends, Python service layers, async media and data pipelines at 6B+ rows, agent frameworks orchestrating multiple model providers. VFX-trained: fire, water and destruction simulations were my tuning threshold.

6B+

rows at sub-500ms P90

25+

ML systems integrated

20+

teams adopted end to end

About me

I am Vansh Bordia, a full-stack engineer from the creative side of the house. I studied VFX (distinction) with a technical-director trajectory before pivoting to AI engineering when generative models arrived: I read the early diffusion papers and predicted the entire creative production pipeline would rebalance on top of them, so I chose to learn to build the machinery rather than master yet another tool release.

The machinery I shipped is real: a platform serving 20+ professional esports organizations at 6B+ rows across PostgreSQL, ClickHouse and BigQuery, with an in-house Python agent harness that orchestrated fine-tuned and hosted models, routed between them, and validated every output before it reached a user. Async pipes rarely broke because I built the validation flows that caught silent failures before customers did. The team that acquired it got the full stack, models included.

Why me

I own features end to end: from the canvas of the interface itself through the async pipeline that keeps it fast, to the agent layer that orchestrates models behind it. Speed is a feature: sub-500ms P90 at 6B+ rows meant real-time analysis felt live, and that is standard I bring to creative tooling too.

I trained in visual effects first, headed toward technical direction (Houdini, Maya, Cinema 4D: fire, water and destruction simulations where craft meets engineering constraint), then pivoted to AI in the diffusion era because I wanted to build the new creative production pipeline, not adapt to it. Since then I have shipped full-stack products solo: React/Next.js and TypeScript frontends, Python service layers, async ingestion pipelines reconciling multi-source feeds into 6B+ analytical rows at sub-500ms P90, and an in-house agent framework orchestrating multiple model providers with RAG, routing, and output validation. My open-source data library has 1,000+ developer downloads, and the platform I built was adopted by 20+ professional esports organizations before its acquisition.

Built and owned the full stack of a platform serving 20+ organizations, through to acquisition

6B+ rows on PostgreSQL, ClickHouse, BigQuery and Redis at sub-500ms P90 query performance

In-house Python agent framework: multi-provider model orchestration, RAG, routing, 25+ output-validating classifier gates

VFX-trained: Houdini, Maya, Cinema 4D, After Effects, with a technical-director trajectory (destruction, fire, water sims)

Open source contributor: data library with 1,000+ developer adopts, SQL tooling, browser image tools

Read the diffusion papers early and pivoted to AI engineering to build the new creative production pipeline

Role stack

TypeScript / React / Next.js / Node / Python / FastAPI / System Design / Distributed Systems / Async Pipelines / Concurrency & Performance / APIs & Integrations / LLM Orchestration / Multi-Model Routing / RAG / Agent Frameworks / Output Validation / MCP Servers / PostgreSQL / ClickHouse / BigQuery / Redis / Apache Spark / Docker / Kubernetes / CI/CD / Observability / Testing / Git / AWS / Houdini / Maya / Cinema 4D / After Effects / Open Source