I’m Sanjay. AI Systems Engineer

I turn ‘what if’ into what’s next.

I design, build and run AI infrastructure, real-time platforms and automation—from first sketch to production.

A compact matte black AI inference server with machined cooling fins and a soft green status light
Availability
Open to architecture, infrastructure & systems work
Open
Local time
India Standard Time, UTC+05:30
  • Python
  • TypeScript
  • FastAPI
  • React
  • Local LLMs
  • Multi-GPU
  • Linux
  • Docker
  • Playwright
  • WebSockets
  • RTMP

What I do

Every layer, one owner.

Architecture, infrastructure, code, deployment and the pager afterwards. Systems fail at the seams between people.

  1. 01AI inference layerlocal models · multi-GPU
  2. 02Real-time mediaingest · transcode · route
  3. 03Automation & agentsplan · execute · verify
  4. 04Backend servicesAPIs · queues · pipelines
  5. 05Interfacesdashboards · control panels
  6. 06Infrastructurelinux · docker · network
  7. 07Operationsmonitoring · recovery

01 / 04

Local inference

Private inference infrastructure, owned outright rather than rented: local models across multi-GPU servers, with task-based routing between them.

Learn more

Selected systems

Built to do real work.

Anonymised systems. Real architecture, decisions and ownership.

Rows of GPU server racks with green status lights in a bright, quiet data room

Multi-GPU · GGUF · Quantization

Distributed AI inference platform

Private, multi-server LLM capacity with task-based model routing behind one internal API.

From idea to production

The whole journey. Owned.

Most technical work stalls between the whiteboard and the server. I stay responsible for the whole line.

  1. 01 / 03 · before any code exists

    Design

    • ProblemWhat actually needs solving, and what does not.
    • ArchitectureThe shape of the system, decided before anything is written.
    • InfrastructureServers, network and storage sized to the real workload.
  2. 02 / 03 · the system itself

    Build

    • BackendServices, APIs and the data model underneath them.
    • InterfaceDashboards and controls someone else can operate safely.
    • AI integrationModels wired in only where they earn their place.
  3. 03 / 03 · and keep it alive

    Operate

    • AutomationRepetitive work removed, with verification built in.
    • MonitoringHealth, alerting and honest failure states.
    • ProductionLive, owned, and still my responsibility afterwards.

How I work

Systems that are owned, verified and operable.

Six things I hold to when deciding how a system should be built, and what to do when the pressure is on to cut a corner.

  • Own it end to endArchitecture, infrastructure, code, deployment and the pager afterwards. Systems fail at the seams between people.
  • Choose boring where it countsTechnology picked on reliability, cost and maintainability, not on how interesting it is to write about.
  • Build for real usersA demo proves an idea works. Production has to survive bad input, partial failure, and someone using it at 3am.
  • Verify, do not assumeAutomation reports what it thinks it did. Independently checking the resulting state is what makes it trustworthy.
  • Design for failure firstFailover, back-pressure and recovery paths go in at the start. They cannot be added convincingly later.
  • Make it operableInfrastructure only one person can run is a liability. The goal is an interface someone else can use safely.

Technical range

The right tools. The bigger picture.

Grouped by what it is used for, not rated out of ten.

Technical capabilities by area
AreaFocusTools and techniques
01Languagesprimary
  • Python
  • TypeScript
  • JavaScript
  • Bash
  • SQL
02Frontendinterfaces
  • React
  • Responsive interfaces
  • Dashboards
  • Real-time UI
  • State management
  • WebSockets
  • Tailwind
03Backendservices
  • FastAPI
  • Node.js
  • REST APIs
  • WebSockets
  • Authentication
  • Background tasks
  • Data pipelines
04AI & ML toolinginference
  • Local LLM deployment
  • GGUF
  • Quantization
  • Agent architecture
  • Model routing
  • GPU optimisation
  • OCR
  • Speech-to-text
  • Vision models
05Automationorchestration
  • Playwright
  • Browser automation
  • Workflow orchestration
  • Event-driven processes
  • Monitoring
  • Error recovery
06Infrastructureplatform
  • Linux
  • Docker
  • Nginx
  • GPU servers
  • Networking
  • DNS & SSL
  • Firewalls
  • RTMP
  • OBS

What are we building next?

Tell me what you’re imagining. We’ll work out what it takes to build it.

  • Architecture reviews
  • AI infrastructure
  • Real-time systems
  • Automation platforms
Let’s discuss your idea