Currently leading a 6-person product team at Kcube AI
I design, build, integrate and lead real software products — from the first requirements call to production on Azure, and increasingly with AI at the core.
Senior Software Engineer / Project Manager
.NET • Azure • Angular • React • AI • Cloud
01 InterfaceAngular and React front ends, typed against the API contract.
01About
For five years I've built production software for organisations that depend on it — from a government reporting portal used across nine districts to recruitment platforms, enterprise AI search and AI media products.
Today I lead a cross-functional team at Kcube AI while staying hands-on in the architecture and the hardest code. I talk to clients, shape requirements, plan the work — and then help build it.
How a problem becomes a product
01
Problem
What's actually broken, for whom
02
Architecture
The simplest shape that holds
03
Development
Typed, tested, reviewable
04
Integration
Third parties, identity, data
05
Deployment
CI/CD, containers, cloud
06
Monitoring
Logs, metrics, real usage
07
Improvement
Evidence-driven iteration
02Experience
From a government portal to enterprise SaaS, AI products and team leadership — each role added a layer: engineering, then architecture, then delivery.
Jun 2025 — Present
Lahore, Pakistan
CurrentSep 2022 — Jun 2025
Lahore, Pakistan
Jan 2021 — Sep 2022
Muzaffarabad, Pakistan
2015 — 2019
University of Azad Jammu and Kashmir · First Division
03Selected work
Five systems I've helped design, build and ship — each one a different kind of hard problem. Open any of them for the full case study.
AI · Marketing intelligence
Turning scattered marketing data into answers a team can act on.
An AI-powered marketing intelligence platform that unifies CRM, analytics, SEO and call data into one warehouse and lets LLMs reason over it with the right context.
Recruitment · Workforce ecosystem
One ecosystem for candidates, staff and clients — and the systems behind them.
A connected set of portals for Agri Labour Australia covering the full candidate lifecycle: applications, onboarding, compliance, payroll documents and client job management.
50,000+daily active users on the Candidate Portal
AI · Video creation
A browser-based editor where AI does the tedious parts of video.
An AI video editing platform with auto-subtitles, AI voiceovers and text-to-video generation, delivered as an installable PWA.
AI · Image processing
Upscale, restore and restyle images with AI — reliably, at any size.
An AI image platform for upscaling, restoration and photo-to-illustration, built around a robust job pipeline for long-running processing.
AI · Enterprise search
Ask your company's documents a question — and get a grounded answer.
An enterprise conversational AI assistant that retrieves from SharePoint with hybrid RAG, respecting who is allowed to see what.
85%+answer accuracy on enterprise conversational search
Also built
Personal project: a URL shortener with FastAPI and Next.js shipped as one Docker image, Postgres + Redis on Azure, and Google OAuth.
FastAPI · Next.js · Redis · Docker
Investor onboarding and property workflows with HubSpot CRM sync and Memberstack authentication.
HubSpot · Memberstack · REST APIs
Scheduled auto-rotation of Google Cloud service-account keys, versioned in Azure Key Vault.
Azure Functions · Key Vault · GCP IAM
RBAC-secured content and directive tracking with KPI dashboards across 9 districts.
ASP.NET · SQL Server
04Capabilities
Not a wall of logos — six areas I work across, and the products where each one shows up.
Typed, well-bounded services that stay readable as they grow.
Where it shows up
05How I work
A product philosophy more than a process — six habits that keep systems simple, honest and ready to change.
“Business problem first, technology second.”
I start with the people and the process — what outcome matters, what constraints are real — before choosing a single tool.
“Reduce the problem to its essential moving parts.”
Most complexity is accidental. I cut scope to what must exist now and name what can wait.
“Designed for reliability, maintainability and change.”
Clear boundaries, explicit contracts and identity designed in from day one, so the system can evolve safely.
“Simple architectures that can evolve.”
I prefer a well-structured monolith or a small set of services over premature distribution — and I stay hands-on in the code.
“Logs, metrics, tests and real user behaviour.”
Assumptions get checked against evidence: automated tests, telemetry and how people actually use the product.
“Optimise only where evidence says it's needed.”
Caching, queues and horizontal scale are tools for measured bottlenecks, not decorations.
06Architecture
Example architecture patterns I work with — an illustration of how I think about layers, boundaries and data flow, not a diagram of any client's proprietary system.
Example architecture patterns I work with
07AI engineering
I treat AI as part of a complete system — data, retrieval, validation and people — not a single API call bolted onto an app.
1// Illustrative — a model call is just a typed dependency2record Answer(string Text, Citation[] Sources);34var context = await search.HybridAsync(query, user.Roles);5var answer = await llm.CompleteAsync<Answer>(prompt, context);67if (!answer.IsValid || answer.Sources.Length == 0)8 return Review.RequestApproval(answer);From data to product experience
Data
Clean, modelled, permissioned sources
Context
Only what the task needs, shaped deliberately
Retrieval
Hybrid vector + keyword search
Reasoning
LLMs with structured outputs
Validation
Schemas, checks and guardrails
Human approval
People stay in control of what matters
Product
Delivered where users already work
↳ 01
Claude, GPT and Gemini integrated as product features with clear failure modes.
↳ 02
Vector, keyword and hybrid retrieval over enterprise content with access control.
↳ 03
Deciding exactly what the model sees — and what it must never see.
↳ 04
Multi-step AI workflows that call tools and hand off to people.
↳ 05
Typed, schema-validated responses the rest of the system can trust.
↳ 06
Prompts treated as versioned, testable code — not magic strings.
08Project management
Alongside engineering, I run delivery: understanding what clients need, planning it, sharing it out across the team, and making sure it ships.
Today
Leading a 6-member team of developers, QA analysts and designers at Kcube AI.
I own requirements and stakeholder communication with international clients, turn goals into roadmaps and estimates, and stay hands-on for architecture decisions and the most complex development.
09Contact