Forward Deployed Engineer @ Global University Systems

Faizullah.

I build AI learning platforms

I build agentic AI systems and AI learning platforms, and run the cloud infrastructure underneath them, embedded with the teams that use them.

See the work Read the blog

GitHub LinkedIn Blogpress ⌘K

livefaiz@ai · shell v3--:--:--

faiz-ai.dev · portfolio shell v3 · type `help`

❯whoami

Faizullah · Forward Deployed Engineer

❯cat now.txt

role Forward Deployed Engineer @ Global University Systems

focus Agentic AI · AI-based LMS platform · forward deployment

stack Python · Google ADK · Strands · Kubernetes · AWS · PostgreSQL

location New Delhi, India

❯

Python
Google ADK
Strands
MCP
LangGraph
RAG
PostgreSQL
Kubernetes
ArgoCD
AWS
Terraform
ECS
GitLab CI
Splunk
Kafka
gRPC
FastAPI
TypeScript
React
vLLM
Python
Google ADK
Strands
MCP
LangGraph
RAG
PostgreSQL
Kubernetes
ArgoCD
AWS
Terraform
ECS
GitLab CI
Splunk
Kafka
gRPC
FastAPI
TypeScript
React
vLLM

01who / what / now

About

summary

Forward Deployed Engineer at Global University Systems with a DevOps and platform background. I design agentic systems for an AI-based learning management system, sit with the product and academic teams who use it, and run agentic-design workshops that turn their workflows into agents, tools and evals. The Kubernetes, AWS and model serving underneath is mine too. Before this I built a multi-tenant agent platform at Kenverse and operated AWS infrastructure at S&P Global across eight environments.

now.txt

role: Forward Deployed Engineer @ Global University Systems
focus: Agentic AI · AI-based LMS platform · forward deployment
stack: Python · Google ADK · Strands · Kubernetes · AWS · PostgreSQL
location: New Delhi, India

Agentic systems

Multi-agent workflows with real tool boundaries, memory, evals and human checkpoints. Shipped on Google ADK, Strands and MCP.

AI learning platform

An AI-based LMS for a global university group: tutoring, course generation and assessment agents grounded in institutional content, built with the people who teach.

Platform

The Kubernetes, AWS, GitOps and observability that keep it running, plus model serving when a workload needs it. Eight environments at S&P Global taught me the boring parts matter.

local time--:--:--

New Delhi, India · IST (UTC+5:30)

open to interesting problems

elsewhere

02where I've shipped

Experience

Forward Deployed Engineer

Global University Systems ↗

Aug 2026 – PresentNew Delhi, India

Embedded with product and academic teams to take AI features from idea to production on GUS's AI-based learning management system.

Design and ship agentic AI systems for the AI-based LMS platform: tutoring, course-generation and assessment agents grounded in institutional content
Build the AI LMS platform end to end with product and academic teams: agent workflows, retrieval over course material, evals and human-in-the-loop checkpoints
Lead agentic-design workshops with business and academic stakeholders, turning real workflows into agent, tool and evaluation designs
Run LLM inference with vLLM where self-hosting is needed: model serving, quantization and autoscaling on GPU nodes

PythonGoogle ADKStrandsKubernetesAWSPostgreSQLRAGvLLM

#AI LMS Platform

Build multi-agent workflows for the LMS: content ingestion, syllabus-aware tutoring, rubric-based grading and learner progress summaries.
Ground agents on institutional material with retrieval pipelines and strict citation of sources.
Define evaluation sets with academic staff and gate releases on eval scores rather than demos.

#Forward Deployment

Sit with the teams using the system, translate their workflows into agent designs and iterate weekly on what ships.
Run hands-on agentic-design workshops covering tool boundaries, memory, human-in-the-loop checkpoints and failure modes.

#Platform & Inference

Deploy and operate the platform on Kubernetes and AWS with observability, guardrails and rollout controls.
Serve open-weight models with vLLM behind an OpenAI-compatible gateway when a workload needs self-hosting.

03what I reach for

Skill map

Agentic AI

Google ADK
Strands
LangGraph
MCP
RAG
Tool design
Memory
Evals
Guardrails

Cloud & Infrastructure

AWS
GCP
Terraform
Ansible
ECS
Fargate
Lambda
S3
Glue

Containers & Delivery

Kubernetes
Docker
ArgoCD
Helm
EKS
GitOps

CI/CD

GitLab CI
GitHub Actions
Azure DevOps
Octopus Deploy
Jenkins
ServiceNow

Observability & Security

Splunk O11y
APM
PagerDuty
Tanium
CrowdStrike
Vault
Consul

Backend & Messaging

FastAPI
gRPC
Express.js
Kafka
RabbitMQ
Redis

Languages & Data

Python
Go
TypeScript
HCL
Shell
PostgreSQL
MongoDB
DuckDB

Inference & Model Serving

vLLM
LiteLLM
Quantization
OpenAI-compatible APIs
Gemini
GPU scheduling

04things I've built

Projects

featured / 01

Jev eSketcher

A generative painting instrument driven by a calibrated decision model

You select part of a sketch; Jev, TypeSafe's System One model, decides which paint material it gets, and the material flies across the desk and paints itself in. 105 procedural sketches, 121 paint materials with their own reveal animations, an infinite canvas engine written from scratch, Play and Sampling modes, and probability fields that show the model's real uncertainty.

TypeScriptCanvasFastAPISQLiteTypeSafe JevDocker

featured / 02

Gigan

Multi-agent runtime that reproduces cloud failures in isolated sandboxes

Turns a problem statement or Terraform project into an isolated AWS-compatible environment where agents inspect, change, validate, pause and replay without touching production. Strands-based agent loop, a policy layer for every action, a Docker control plane and a durable evidence store in PostgreSQL.

PythonStrandsFastAPIPostgreSQLDockerFloci

featured / 03

F433

Autonomous AI football social network with live data

An autonomous social network where AI analysts debate, predict and react to football events in continuous shifts. Parallel shift engine on Google ADK, FastAPI and PostgreSQL with dynamic skill injection and live API-Football data.

FastAPIReactPostgreSQLGoogle ADKLiteLLM

🏅 Winner, Smart India Hackathon 2023

Medusa AI Proctoring

Real-time AI exam proctoring system

ReactNode.jsWebRTCGemini

open source

codestash

FastAPI + React starter with an AI agent generator

FastAPIReactPostgreSQLTerraform

freelance

Sacred Shaadi

Matrimonial platform: backend and infrastructure

Node.jsTypeScriptAWSPostgreSQL

freelance

FozDeal

Multi-vendor e-commerce marketplace: backend and cloud

Node.jsTypeScriptAWSMongoDB

open source

GenAI Fashion Outfit Generator

Outfit generation with chat-based style advice

DjangoPostgreSQLGeminiStable Diffusion

open source

Alphamart

Full-featured e-commerce platform

DjangoDRFReactRedux

05notes from production

Writing

All posts →

Aug 5, 2026·11 min read

Gigan: Turning Cloud Failures into Reproducible Evidence

Inside the architecture of a multi-agent runtime that recreates cloud failures in isolated Floci sandboxes and verifies each repair.

Agentic AIDevOpsAWSSRE

May 3, 2026·4 min read

The Salesforce Agent That Actually Works in Production

Converting natural language into SOQL looks simple in a demo and breaks in predictable ways in production. The system-level decisions that made it reliable.

Agentic AILLMSalesforceSOQL

06education / awards / certs

Background

Education

Jamia Millia Islamia

B.Tech in Computer Science

2020 – 2024 · New Delhi, India · CGPA 8.7

Awards

SIHSmart India Hackathon 2023

Winner of the national hackathon organized by the Government of India, for the Medusa AI proctoring system.

LCLeetCode 500+

Solved over 500 algorithmic problems.

Certifications

07open to interesting problems

Contact

Building something with agents, an AI learning product, or the infrastructure under them?

I reply to every real message. Forward-deployed work, platform consulting, or a second opinion on an architecture.

Email me

GitHub ↗LinkedIn ↗Blog ↗LeetCode ↗