8000 sunilgawai (Sunil Gawai) · GitHub
[go: up one dir, main page]

Skip to content
View sunilgawai's full-sized avatar
💘
Dating React Native
💘
Dating React Native

Block or report sunilgawai

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
< 8000 /dialog>
sunilgawai/README.md

Typing SVG


$ whoami

class SunilGawai(AppliedAIEngineer):
    """Backend engineer building the production layer that AI agents run on."""

    title      = "Applied AI Engineer · Agentic Systems"
    location   = "Pune, India 🇮🇳"
    experience = "2+ years in production · Full Stack Engineer @ Swift Business Solutions · ex-Wipro"
    stack      = ["TypeScript", "Node.js", "Python", "PostgreSQL", "Redis", "RabbitMQ", "Docker"]

    shipped = [
        "Stellar POS: 0→1 real-time retail platform across multiple store locations",
        "Multi-tenant operations platform serving 3 companies on shared infrastructure",
        "MCP server that lets Claude, Gemini and ChatGPT operate on a live codebase",
        "Transformer Atlas: interactive, tested model of attention and causal masking",
    ]

    building = ["LLM pipelines", "AI agent architectures", "MCP tooling", "OpenTelemetry", "Rust"]

    def why_me(self):
        return (
            "An agent is a distributed system with a language model inside it. "
            "Queues, retries, tenant isolation, access control and observability "
            "are what make it reliable, and that's what I've shipped in production."
        )

    def contact(self):
        return "sunilgawaiofficial@gmail.com"

🧠 What I build

Area What that means in practice
🤖 Agent tooling with MCP Model Context Protocol servers that give AI clients real tools: insert, search, list and remove code in a workspace
⚙️ Async pipelines for long-running work RabbitMQ and Redis pipelines that decouple producers from consumers, the backbone agent jobs need to survive load and failure
🔐 Guardrails before AI touches data Multi-tenant systems with tenant-level isolation and company-scoped RBAC, so every action stays inside its access boundary
⚡ High-throughput APIs Node.js and TypeScript REST APIs, tuned PostgreSQL schemas and query plans, Redis caching on hot paths
🔗 Event-driven integrations Webhook ingestion from Shopify, WooCommerce and Global Payments, processed through queues instead of inline
🧮 Model internals Attention, softmax normalisation and causal masking implemented and unit-tested, not just called through an API

🚀 AI projects

An interactive learning studio that walks through the full Transformer architecture in 20 chapters. Tokens, attention maps, vectors and generation controls all stay in sync with one playback clock.

  • Deterministic attention maths with causal masking
  • Unit tests for the maths, Playwright tests for the UI

🛠️ Deep MCP Logger

A VS Code extension with a built-in MCP server. AI clients like Claude, Gemini and ChatGPT can insert, search, list and clean up log statements across a workspace.

  • AST-aware edits across JS, TS, Python, PHP and Go
  • Standalone MCP server with six tools

A conversational AI guide built on the Inworld AI character engine. Users chat with it in real time, with history, authentication and per-user API rate limits.

  • Chat window with saved history
  • NextAuth and Prisma for accounts, with rate limiting on the AI API

💬 Veyra: conversation-to-order platform

A multi-tenant B2B platform that turns WhatsApp messages like "20 case Parle-G, 10 Marie, Thursday" into priced, stock-checked, credit-checked orders and GST invoices.

  • Pipeline: conversation → understanding → domain operation → event → automation
  • 2,600+ tests across 13 packages, every package at 90%+ line coverage

🛠️ Tech stack

AI & agents

Backend, data & messaging

Frontend & mobile

🔁 How I build an agent for production

flowchart LR
    U[User message<br/>or event] --> A[LLM agent]
    A -->|tool calls| T[MCP tool layer]
    T --> Q[(Queue<br/>RabbitMQ)]
    Q --> S[Domain services<br/>Node.js + TS]
    S --> D[(PostgreSQL<br/>+ Redis)]
    G[RBAC and<br/>tenant isolation] -.-> T
    O[Observability<br/>OpenTelemetry] -.-> S
    S -->|events| A
Loading

📊 GitHub activity

GitHub stats Top languages

GitHub streak

🤝 Let's build something intelligent

I'm open to Applied AI Engineer roles, especially on teams building agentic systems and taking LLM features from demo to production.

Pinned Loading

  1. morpha morpha Public

    Framework‑agnostic engine for visual docs — power slide editors, pitch decks, whiteboards, infographics, and resumes. External formats are translations, never the model.

    TypeScript

  2. transformer-atlas transformer-atlas Public

    Interactive Transformer Architecture

    TypeScript

  3. www.pizzapoint.com www.pizzapoint.com Public

    TypeScript

  4. www.foodie.com--server www.foodie.com--server Public

    This is a Node/typescript server. A server for ecommerce based online food ordering site, with functionality of Real Time Order Status using Socket, Advance Seat Booking and many more.

    TypeScript 3

0