Skip to content

span 01 · hero · sunil gawai · pune

I build the systemsAI agents run on.

Backend engineer turned applied AI engineer — queues, pipelines and multi-tenant systems, now powering reliable LLM agents.

View work
  • TOOL USE
  • AGENT LOOPS
  • RAG
  • MCP
  • EVALS
  • STREAMING
  • ANTHROPIC SDK
  • NODE.JS
  • TYPESCRIPT
  • EXPRESS
  • PYTHON
  • REST API DESIGN
  • POSTGRESQL
  • MYSQL
  • MONGODB
  • REDIS
  • RABBITMQ
  • DOCKER
  • CI/CD
  • OPENTELEMETRY (EXPLORING)
  • MULTI-TENANCY
  • RBAC
  • REACT
  • NEXT.JS
  • VUE.JS

span 02 · positioning

I build the queues , retries , memory and guardrails that let AI agents run in production — the unglamorous half of every demo that actually ships.

span 04 · anatomy of an agent run · 7 spans

Every agent run is a trace. Mine come with retries.

Scroll through one run of the agent on this site. Each span maps to backend work I've already shipped.

  1. 01 · user.prompt · 0–120msInput validation & limitsSchema-validated request bodies and hard limits on every public API I've shipped.
  2. 02 · agent.plan · 120–520msDecomposing work into stepsDesigned multi-module platforms end to end — Stellar POS and the Diamond Trading platform.
  3. 03 · tool.search_projects · 520–900msTenant-scoped retrievalCompany-scoped RBAC and data isolation for three companies on one platform.
  4. 04 · tool.get_experience · 560–980msParallel, idempotent tool callsQueue-backed webhook ingestion for Shopify, WooCommerce and Global Payments at Stellar POS.
  5. 05 · model.generate · 980–1900msStreaming & backpressureRabbitMQ pipelines that decouple producers from consumers at Swift Business Solutions.
  6. 06 · guardrail.check · 1900–2150msGuardrails & access controlRole-based permission systems running in production.
  7. 07 · agent.answer · 2150–2400msObservabilityObservability hooks and production rollout ownership on the Diamond Trading platform.

Capabilities

Agents & LLMs
Tool use, Agent loops, RAG, MCP, Evals, Streaming, Anthropic SDK
Backend
Node.js, TypeScript, Express, Python, REST API design
Data & messaging
PostgreSQL, MySQL, MongoDB, Redis, RabbitMQ
Infra & reliability
Docker, CI/CD, OpenTelemetry (exploring), Multi-tenancy, RBAC
Frontend
React, Next.js, Vue.js

span 05 · experience · 4.8y

Experience, as a trace

  1. 2023·06 → now · 3.4y

    Full Stack Engineer

    Swift Business Solutions · Remote

    • Designed REST APIs in Node.js + TypeScript for concurrent production traffic; tuned PostgreSQL schemas and query plans on high-throughput paths.
    • Built RabbitMQ-backed async pipelines that decouple producers from consumers for resilient background processing.
    • Ran containerised services on Docker with Redis caching on hot read paths to cut database pressure under load.
    • Owned backend architecture across POS and e-commerce platforms — data models, API contracts and performance.

    child spansStellar POS · Diamond Trading Operations Platform

    Node.jsTypeScriptPostgreSQLRabbitMQRedisDockerVue.jsReact

  2. 2022·10 → 2023·04 · 7mo

    Software Engineer

    Wipro · Pune, India

    • Contributed to enterprise backend systems: large-scale distributed service workflows, cross-team delivery and production code reviews.
  3. 2022·01 → 2022·09 · 9mo

    Freelance Full Stack Developer

    Self-employed · Remote

    • Built custom web applications for small businesses with React, Node.js and MySQL.
    • Delivered end to end — requirements, build and deployment — earning repeat contracts.

    ReactNode.jsMySQL

span 06 · contact

Building agents that need to survive production?

sunilgawai7397@gmail.com
GitHubResume