P—M

AI ARCHITECT · AUTODESK

Prakash MaheshwaranNew York · Building globally

I make
intelligence
useful.

I architect AI systems, automate stubborn work, and engineer products that hold up after the demo ends.

Right now

Architecting production AI at Autodesk Building Flint in the open Shipping practical tools for n8n
Meet the person behind the systems
SCROLL TO ENTER ↓
AI ARCHITECTURE ✦AGENTIC AUTOMATION ✦PRODUCT ENGINEERING ✦HUMAN SYSTEMS ✦OPEN SOURCE ✦AI ARCHITECTURE ✦AGENTIC AUTOMATION ✦PRODUCT ENGINEERING ✦HUMAN SYSTEMS ✦OPEN SOURCE ✦

01 / HELLO

Animated illustration of Prakash waving helloA tiny hello from Kash

Most people call me Kash.

Tamil Nadu raised.
New York based.
Still curious.

I'm Prakash—an AI architect at Autodesk, an open-source builder, and the kind of person who keeps pulling at a problem until the demo becomes something people can actually rely on.

I grew up in Tamil Nadu, started in computer science in 2019, and now work from Binghamton, New York. The thread through all of it is simple: I like making complicated technology feel useful, honest, and human.

01 · Builder before title02 · Open source by instinct03 · Gym keeps the brain honest

02 / POINT OF VIEW

The interesting part of AI is no longer the demo.

It is everything that makes the demo survive reality.

My work lives in that gap: between a powerful model and a system people can depend on.

03 / THINGS I'VE BUILT

Built because I wanted them to exist.

A deliberately curated set—not an automated ranking of whichever repository has the most stars.

01Kotlin + SwiftJul 2026

Flint — Peak Focus

My most ambitious open-source product: native iOS and Android focus tools with no paywall, accounts, or telemetry. The entire blocking engine stays free.

KotlinSwiftUIScreen TimeLocal-first
Source ↗Visit ↗
02TypeScriptJul 2026

n8n FFmpeg WASM

A pure-WebAssembly media workstation inside n8n: 21 video and audio operations, with no native FFmpeg binary or custom Docker image required.

n8nFFmpegWebAssembly21 operations
Source ↗Visit ↗
03TypeScriptJul 2026

n8n YouTube DL

A plug-and-play n8n node for video, audio, transcripts, and subtitles across 1,000+ sites—even inside Alpine-based Docker environments.

n8nyt-dlpTranscripts1,000+ sites
Source ↗Visit ↗
04Python + VueNov 2025

Moplexity

My answer to closed AI search: multi-source web, Reddit, and YouTube research with citations, model choice, and local conversation history.

FastAPILiteLLMCitationsOpen source
Source ↗
05PythonMay 2026

AVANGRID P75 Framework

Winner of the 2025 Advanced Analytics & Risk Management category: a renewable-energy pricing framework using GARCH, Monte Carlo simulation, and ML forecasting.

Hackathon winnerGARCHMonte CarloEnergy
Source ↗
06TypeScriptMay 2026

n8n LaTeX to PDF

Document automation for n8n with three TeX engines, custom packages, multi-pass compilation, bibliographies, and binary-data workflows.

n8nLaTeXPDFDocument automation
Source ↗Visit ↗
Explore the full GitHub archive↗

04 / THE ROAD HERE

The recent work,
and everything
that led to it.

2026—NOW

AI Architect, People & Places

Autodesk

Designing agentic and LLM systems for enterprise use—especially the validation, ownership, and architectural layers that determine whether production AI can be trusted.

2025—26

Senior Research Aide

The State University of New York

Turned applied AI research into production-minded prototypes, with a focus on MLOps, automation, and systems that could leave the lab.

2025

Data Engineer & AI Automation Architect

Flomenco

Built digital automation and data systems for media workflows, connecting creative operations with dependable technical infrastructure.

2025

AI Developer

Skyvern · Y Combinator S23

Worked on browser agents that had to survive slow pages, shifting interfaces, and the ambiguity of real enterprise workflows.

THE EARLIER CHAPTERS

The compact version—because the path matters, even when every stop does not need a paragraph.

2024—26

M.S. Computer Science · Binghamton University

2023—24

M.S. Computer Science, AI track · VIT

2023

Cloud & DevOps Engineering Intern · uniqe technologies

2019—23

Computer Science · Anna University

2019—21

Volunteer Educator · U&I Trust

05 / NOTES TO SELF

Notes from the workbench.

Ideas about intelligent systems, automation, and making technology more human.

01

Building AI systems people can actually trust

A practical view of evaluation, human oversight, and the unglamorous work between a model and a product.

6 MIN READ↗02

Browser agents beyond the perfect demo

What changes when automation meets slow pages, ambiguous interfaces, and real operational risk.

5 MIN READ↗03

Open source as a way of thinking

Why building in public sharpens product decisions, technical judgment, and the quality of the work.

4 MIN READ↗

06 / NEXT

Say hello.
I'm listening.

No pitch deck required. If something here sparked an idea—or you just want to compare notes about agents, open source, or building in public—send me the messy version.