I build things that shouldn't exist yet — then I sell them.

I'm Vigneshwar — a CS student at Chennai running three careers simultaneously: systems engineer, ML researcher, and digital entrepreneur. While most students are submitting assignments, I'm shipping products to real customers on Reddit, growing communities on Discord, and inventing neural architectures from my laptop.

My primary language is Python — it's where I'm strongest, fastest, and most creative. I use it for everything from ML research to backend APIs to automation pipelines. I also work with C and C++ for systems-level work (built an 8-world game engine with C++17 + Raylib), and across the full web stack in TypeScript, JavaScript, and GLSL for GPU shaders.

I've built and sold 7 digital products — eLearning platforms, developer tools, automation systems, analytics dashboards — all launched organically through communities. No VC, no paid ads, no gatekeepers.

The combination is the edge: I can write the code, design the architecture, train the model, write the copy, run the campaign, and close the sale. Most people pick one. I run all of them and they compound.

Primary: Python · FastAPI · PyTorch · React/Vue · Three.js · AWS/GCP
Systems: C++17 · Raylib · CMake · WebAssembly · GLSL
Currently: RAG system architecture · Two Python libraries · NEURALSTACK (stealth) · Shipping daily

Python Dev My primary stack. ML pipelines · FastAPI · automation · data tools · CLI libraries. Where I'm fastest.

ML Architect Novel transformer architectures. Graph topology · temporal decay · adaptive compute. Not fine-tuning — designing.

WebGL Wizard Three.js · GLSL shaders · WebAssembly · 600K GPU particles in-browser. Pushing what browsers can do.

Growth Hacker Google Ads · Meta · SEO · Reddit launches · Discord communities. Zero ad spend, organic conversions.

Entrepreneur 7 digital products built and sold. eLearning · dev tools · automation. Different platforms, different prices.

Cloud Optimizer AWS + GCP cost optimization. Avg 30-50% infra reduction. Grafana monitoring. Startups love this.

The eLearning Platform — Sold for 68K

Built a complete eLearning platform from the ground up: course infrastructure, payment integration, student progress tracking, instructor dashboards, content delivery system. Marketed and sold exclusively through organic community channels — Reddit threads, Discord servers, and developer forums. Found its buyer without a single rupee spent on advertising.

The Playbook:

STEP 1 | Identify real pain in a niche technical community
STEP 2 | Build the solution fast, focused, polished
STEP 3 | Soft-launch on Reddit - genuine posts, no hype
STEP 4 | Engage authentically - answer questions, give value
STEP 5 | Let the product speak - demos, GIFs, live links
STEP 6 | Collect feedback, iterate, close buyers in DMs
STEP 7 | Repeat with a new product in a new niche

Product Portfolio

Launch platforms: Reddit (primary) · Discord servers · Slack workspaces · Developer forums

Building a production-grade RAG system from scratch — not a wrapper, an architectural contribution. Built on the TemporalMesh Transformer backbone with custom retrieval that understands temporal semantic relationships, not just cosine similarity to static embeddings.

STANDARD RAG:   Query -> Embed -> Top-K -> Stuff context -> Generate
                Problem: No temporal awareness. Static retrieval.

VIGNESHWAR RAG: Query -> Temporal Embed -> Mesh Graph Retrieval
                -> Confidence-gated Context -> Adaptive Generation
                Result: Retrieval that understands WHEN info matters

Library 1 — ML Experiment Tooling

Solves friction in ML experiment management that W&B and MLflow don't handle well. Lightweight, zero config, plugs into any training loop.

# Before: 30 lines of boilerplate per experiment
# After:
@track.experiment(name="tmt-ablation-v3")
def train(config):
    ...  # your existing loop, unchanged

Library 2 — Data Pipeline Utilities

Streaming ingestion, schema validation, temporal windowing, pipeline composition. The abstractions every ML project reinvents from scratch.

pipe = Pipeline([
    Stream.from_api(url, interval=60),
    Window.temporal(size="1h", decay=0.95),
    validate(schema),
])

Release: GitHub + PyPI · Announced on Reddit and Discord

+====================================================================================+
|                     VIGNESHWAR // ACTIVE MISSIONS 2025                             |
+====================================================================================+
|  [DONE] Claude Code Best Practices  > Top production guide to AI-assisted dev     |
|  [DONE] TemporalMesh Transformer    > Novel arch - graph + time + compute         |
|  [DONE] Game of DSA                 > C++ 8-world game · WASM · runs in browser  |
|  [DONE] NEXUS Cosmic                > 600K GPU particles + live news universe     |
|  [DONE] CrypticBastion              > Real-time global cyber threat intel         |
|  [DONE] GITGALAXY                   > 10-planet gamified engineering universe     |
|  [DONE] AttendanceAI                > AI attendance intelligence platform         |
|  [SOLD] Digital Products x7         > eLearning · tools · SaaS · automation      |
+------------------------------------------------------------------------------------+
|  [WIP]  NEURALSTACK                 > AI DevOps intelligence platform [STEALTH]   |
|  [WIP]  RAG System Architecture     > TMT-backed temporal semantic retrieval      |
|  [WIP]  Python Library x2           > ML tooling + data pipeline utils            |
|  [WIP]  Digital Brand Building      > Reddit · Discord · Slack community          |
+------------------------------------------------------------------------------------+
|  PRIMARY: Python · PyTorch · FastAPI · React · Vue · Three.js · AWS               |
|  SYSTEMS: C++17 · Raylib · CMake · WebAssembly · GLSL                             |
|  OPS:     Aerospike · Docker · Grafana · CloudWatch · AWS cost optimization       |
|  MKT:     Google Ads · Meta · SEO · Reddit launches · Discord growth              |
|  EDGE:    Only person who can BUILD it + SELL it + SCALE it + MARKET it           |
+====================================================================================+

I don't launch behind paywalls or in isolation. I build in public, share progress, and let the community validate before I optimize. Free launches first, real feedback always, paid version after trust is built.

Why Reddit works: Technical communities have the highest-intent buyers on the internet. They evaluate rigorously, tell their networks when something is good, and prefer buying from builders they've seen in the community.