Experience

Backend Lead Engineer · MicroAgility Services (Pvt) Ltd., Islamabad

Jan 2024 – Present

Built the AI gateway for Talentnix, a multi-tenant SaaS platform: one routing layer in front of OpenAI, Gemini and Anthropic with automatic failover, which cut AI integration complexity by 60%.
Designed Talentnix’s three-tier resilience pattern (Retry → RAG Fallback → Rule-Based Floor). Users have seen zero AI failures during live LLM outages; the design is written up in the “Designing for AI Failure” case study.
Enforced tenant isolation across Talentnix’s REST and GraphQL APIs with OAuth 2.0/JWT, with no cross-tenant data exposure incidents to date.
Lead backend development for the APAC Management System, MicroAgility’s HRMS that combines payroll, HR and attendance with complete employee records, and for client projects.
Brought dashboard response times down from 12s to under 2s (83%) through PostgreSQL and MongoDB query tuning, new indexes and a MeiliSearch rollout.
Built an event-driven BullMQ and Redis job pipeline that keeps APIs under a second while thousands of async jobs run concurrently.
Delivered full Jest unit-test coverage with GitHub Copilot, and built an internal ATS with Claude Code and Gemini 2.5 Flash.
Set up GitHub Actions CI/CD pipelines with automated test gates, shortening release cycles.
Mentor three associate backend developers through the company’s Learning & Development programme.

Backend Software Engineer · MicroAgility Services (Pvt) Ltd., Islamabad

Mar 2022 – Jan 2024

Owned backend development for iAgility and AgileBrains, two large products built around the hiring and consultancy workflow, on a MEAN (MongoDB, Express.js, Angular and Node.js) microservices architecture.
Built Express.js microservices and GraphQL APIs that handle each product’s hiring flow, with MongoDB as the main data store.
Wired up Postal mail for transactional email and Google Cloud Storage for files, with WebSocket for real-time notifications.
Ran zero-downtime schema migrations on live multi-tenant production databases while maintaining 99.9% uptime.
Refactored the REST and GraphQL services that serve thousands of users across multiple tenants, improving data retrieval speed by about 40%.
Automated the deployment and migration pipeline for new enterprise tenants, cutting manual onboarding effort.

Software Engineer · Market Icon, Rawalpindi

Aug 2018 – Mar 2022

Worked as the primary full-stack developer on a large-scale MEAN (MongoDB, Express.js, Angular and Node.js) system, with ownership of backend design and delivery.
Designed the distributed server and database architecture on an OpenStack cloud platform.
Modelled the databases for continuously growing real-time data, so reads and writes stayed fast as volume rose.
Built serverless AWS Lambda and API Gateway services on an event-driven design, removing the idle cost of always-on servers.
Built Express.js, FastAPI and Django REST Framework APIs for secure CRM and payment integrations.
Hardened backend services against peak load and power outages, and introduced code documentation and change tracking.

Selected work

Talentnix: AI gateway & resilience (Backend Lead): Multi-tenant SaaS platform. One routing layer in front of OpenAI, Gemini and Anthropic with automatic failover, backed by a three-tier fallback (retry → RAG fallback → rule-based floor). Tenant isolation is enforced across its REST and GraphQL APIs with OAuth 2.0 and JWT. OpenAI · Gemini · Anthropic · REST · GraphQL · OAuth 2.0 / JWT

Multi-tenant AI recruitment ATS (Backend Lead & Architect): Internal ATS built with Claude Code and Gemini 2.5 Flash. Tenant scope comes from verified JWT claims across the query layer. Résumé ingestion runs on BullMQ workers with rule-based fallbacks; search is tenant-scoped MeiliSearch. NestJS · MongoDB · Gemini 2.5 Flash · MeiliSearch · BullMQ · Postal SMTP

APAC Management System (HRMS) (Backend Lead): MicroAgility’s HRMS, combining payroll, HR and attendance with complete employee records. It has role-based access control, leave banking with Excel import and export, multi-level request approvals, and GCS signed-URL documents behind a Redis cache. NestJS 11 · MongoDB · Redis · GCS

iAgility (Backend Engineer): A large product built around the hiring and consultancy workflow, on a MEAN microservices stack. I owned the backend: Express.js microservices and GraphQL APIs for the hiring flow with MongoDB as the main data store, Postal for transactional email, Google Cloud Storage for files and WebSocket for real-time notifications. MongoDB · Express.js · Angular · Node.js · GraphQL · WebSocket · Postal

AgileBrains (Backend Engineer): A single consultant and client hiring platform on the same MEAN microservices stack: Admin, Client, Consultant and Job services behind a Main App gateway and a custom service registry, each split into presentation, business and data layers. MongoDB · Express.js · Angular · Node.js · TypeScript · Mongoose

Serverless gateway & CRM layer (Software Engineer): Serverless AWS Lambda and API Gateway services on an event-driven design, replacing always-on servers, plus Express.js, FastAPI and Django REST Framework APIs for secure CRM and payment integrations. AWS Lambda · API Gateway · Express.js · FastAPI · Django REST

Stack

Architecture & systems design: Distributed systems, system design, microservices, event-driven architecture, multi-tenant SaaS, REST and GraphQL API design, scalability and reliability trade-offs

Backend engineering: Node.js, NestJS, TypeScript, Express.js, Python, FastAPI, Django REST Framework, PostgreSQL, MySQL, MongoDB, Redis, BullMQ, MeiliSearch

AI-first engineering: LLM integration (OpenAI, Gemini, Anthropic), RAG pipelines, MCP servers, AI agent workflows, Claude Code, GitHub Copilot, Gemini 2.5 Flash

Cloud & delivery: AWS (Lambda, API Gateway), GCP, OpenStack, Docker, Kubernetes, CI/CD (GitHub Actions), Jest, OAuth 2.0 / JWT

Product domains: Multi-tenant SaaS, HRMS, CRM, recruitment and consultancy platforms

Leadership: Technical leadership, mentoring and onboarding, architecture and code review, technical direction, architecture decision records (ADRs)

Certifications & Credentials

74 Verified Credentials: LinkedIn Learning 23 · Google Cloud Skills Boost 21 · IBM 14 · Anthropic 7 · Microsoft Learning 6 · Udemy 1 · Coursera 1 · Cognitive Class 1. View all

Projects

Omni.io: Multi-tenant RAG support engine with a three-tier resilience ladder, Postgres row-level-security isolation, an embeddable widget and an MCP server. 81 unit and 16 end-to-end tests.

TalntFlow AI: Lightweight ATS with a Kanban pipeline, STAR scorecards, hiring analytics and an approval-based recruiting agent that runs on Claude with a rules fallback.

Scanwise: Formerly Qeloma OCR. Turns scans, photos, PDFs, Office files and zips into one cited summary and a source-grounded chat, with on-device OCR and per-word confidence.

Catchbox: Human-in-the-loop triage for failed web scrapes, with a no-code scraper builder, 30-day run history and Google Sheets sync. Rebuilt in plain language from Scrapefix.

Scrapefix: The first build of the scraper-triage dashboard: fix queue, record review, clean-data warehouse and a personal ingest key for any scraper.

TRIA.GE: Support-triage console where Gemini classifies intent, sentiment and urgency, extracts entities, then resolves tier-1 tickets or escalates to tier-2 with an audit log. Ticket sources and UiPath robot runs are simulated.

SAP-AGI Procure: Procurement and accounts-payable control centre: quotation comparison, 3-way PO, goods-receipt and invoice matching, duplicate-payment flags and vendor scorecards. The SAP, UiPath and Document AI connections are simulated.

GitHub Glow-Up: Lovable-built workspace for repository documentation: drafts READMEs, builds code reports that keep verified public facts apart from the AI's interpretation, and publishes the README, docs, topics and a release to GitHub, then re-reads the repo to confirm.

Taskmaster Agent: Hackathon entry: an event-driven HR operations agent on Google Genkit and Gemini, run through Pub/Sub on Cloud Run. It audits timesheets against policy, pauses for a manager's approval when it finds a violation such as overtime over 10 hours, and indexes every decision in Meilisearch.

Verdict: Tamper-evident decision engine that issues reasoning receipts, hash-chained and Ed25519-signed, built for EU AI Act Articles 12 to 14 record-keeping.

Lens Studio: Capability-driven document understanding service: PDFs, DOCX and images become one canonical envelope, then seven capabilities (summarise, extract facts, verdict, compare, breakdown, next actions, generate) run on Gemini with deterministic fallbacks and API-key tenancy.

Voice Studio: Real-time voice analyst on the Gemini Live API. It calls Lens capabilities as tools, so every spoken answer is grounded in your document and shown in an on-screen trail.

Shift: Semantic diffing for contracts, policies and configs that aligns reworded clauses, ranks changes by materiality and explains business impact.

Cover Studio: Browser-based LinkedIn banner designer with an AI copywriting assistant, safe-zone guides and one-click high-resolution export. Renders client-side.

Qeloma Studio: In-browser photo and video editor on WebGPU, WebGL2 and WebAssembly, with local ONNX background removal and a non-destructive edit graph. No server calls.

Qeloma Recruit: Angular 21 ATS sandbox on Gemini: job-description generator, 10-point STAR interview guides, scorecards, candidate matching, a five-stage pipeline, analytics and PDF export.

Room Booking: Multi-tenant meeting-room booking on NestJS 11 and React 18, with conflict detection, calendar integration, notifications and English, Chinese and Japanese UI.

Extension suite: Ten single-purpose Manifest V3 extensions built with WXT: one-click and full-page screenshots, a domain blocker with Pomodoro timer, zero-flicker dark mode, Gemini-powered page clips, a per-tab volume booster, time-boxing, private time tracking, an SEO inspector, reader mode and a colour palette tool. Local-first, with a Chrome Web Store preflight gate.

Services & Contact

Backend architecture: Multi-tenant SaaS design with tenant isolation, schema strategy and REST or GraphQL API contracts.

AI platform & resilience: Provider-agnostic LLM layers, RAG pipelines and fallback ladders that keep AI features up through outages.

Performance & scale: Query and index work, search, and queue-based workloads that keep APIs sub-second under load.

Technical leadership: Mentoring and onboarding, architecture and code review, and architecture decision records (ADRs).

Islamabad, Pakistan · Open to relocation

Jawad Ul Hadi

Backend Lead / Architect · AI-first systems design

Eight-plus years building multi-tenant SaaS backends in NestJS, TypeScript and PostgreSQL. I design AI features to fail predictably: a three-tier fallback of retry, retrieval and a rule-based floor keeps them working through provider outages.

Work with me Read the case study

years on production backends

83%

dashboard response times cut from 12s to under 2s (83%)

Zero

AI failures during live LLM outages

60%

cut in AI integration complexity from one gateway over OpenAI, Gemini and Anthropic

Case Study Designing for AI Failure Retry → RAG Fallback → Rule-Based Floor: how Talentnix kept zero AI failures during live LLM outages. Read it →

Experience

Career spans · 3 roles on a shared month scale

Career ribbon · time split by role

Selected work

Client systems are under NDA. Architecture is described without business data or endpoints.

Stack & credentials

Education B.S. Computer Science, Government College University, Faisalabad, 2018

Certifications

The five on my résumé, from 74 verified credentials. Each links to the issuer's verification page.

View all 74 credentials →

GitHub activity

My contribution calendar in 3D, regenerated daily from GitHub.

3D GitHub contribution calendar for JawadulHadi, with yearly contribution, star and fork counts

Projects & open source

Personal builds, public and verifiable.

Public repositories

{{ r.name }} {{ r.desc }} {{ r.kind }} ↗

The Qeloma suite

{{ q.name }} {{ q.desc }} {{ q.kind }} ↗

Case study

A full write-up with architecture figures and a failure simulator.

Case study · Resilience pattern

Designing for AI Failure

One abstraction sits in front of OpenAI, Gemini and Anthropic. When a provider degrades, requests step down through three tiers instead of surfacing an error. The gateway cut AI integration complexity by 60%, and users have seen zero AI failures during live LLM outages.

Request

Provider gateway

OpenAI · Gemini · Anthropic

Tier 1

Retry & failover

Backoff, then the next provider

Tier 2

RAG fallback

Answer from retrieved context

Tier 3

Rule-based floor

Deterministic, never hard-fails

Read the full case study →

Services & contact

Let's build something resilient together.

Open to Backend Lead, Solutions Architecture and AI Platform roles: remote from Islamabad with hours aligned to EST/PST, hybrid, on-site, or relocating with support. Every way to reach me lives on one page.

Contact me on Gravatar ↗

Experience

Backend Lead Engineer · MicroAgility Services (Pvt) Ltd., Islamabad

Backend Software Engineer · MicroAgility Services (Pvt) Ltd., Islamabad

Software Engineer · Market Icon, Rawalpindi

Selected work

Stack

Certifications & Credentials

Projects

Services & Contact

Jawad Ul Hadi

Experience

{{ j.role }}

Selected work

{{ w.title }}

Stack & credentials

{{ s.group }}

Certifications

GitHub activity

Projects & open source

Public repositories

The Qeloma suite

Case study

Designing for AI Failure

Services & contact

{{ s.t }}

Let's build something resilient together.