Ahmad Igbaryya
Summary
I enjoy taking an unclear idea and turning it into something simple, useful, and reliable. I like understanding the full picture, asking the right questions, and working closely with people to find a practical way forward.
Selected Achievements
- Raised a telecom data-sync success rate from ≈30% to ≈99% with ~10x throughput (≈50 → ≈500 records/sec) on a platform I built end-to-end.
- Architected a GraphQL federation platform and the standards multiple feature teams built on, cutting new-service bootstrap from days to minutes.
- Built internal tooling in production use by 200+ engineers across 10+ delivery projects, 7+ years and counting.
- Led and mentored a Scrum team, and repeatedly parachuted into at-risk telecom programs across European carriers to stabilize delivery.
- Shipped a youth crisis-management system to live production in 4 days during the October 2023 emergency.
Experience
- Built an AI agent framework (LangChain over OpenAI / AWS Bedrock / vLLM) with streaming chat and 25+ tool integrations, plus a RAG knowledge-base system on PostgreSQL/pgvector (multi-format ingestion, embeddings, semantic retrieval).
- Implemented agent interoperability via the A2A v1.0 protocol (agent cards, registry, JSON-RPC task lifecycle) and built an MCP server exposing 49 platform operations to IDE/AI clients.
- Developed a React/Redux visual workflow builder (20+ step types with nested flow control) over a 50+ endpoint Node.js/Express API with JWT/LDAP auth and RBAC.
- Owned deployment with Helm/Kubernetes (StatefulSets, HPA autoscaling, ingress, migration init containers) and Jenkins CI/CD to a Nexus registry, instrumented with OpenTelemetry tracing.
- Sole engineer responsible for architecture and implementation (backend, frontend, data, CI/CD) of a telecom operations platform automating customer billing-account (BAN) refresh and cross-system data synchronization, working with operations teams and external billing-system owners; lifted the core sync KPI from ≈30% to ≈99% in 10 months and sustained a >95% success rate.
- Scaled a single-threaded Node.js service to process ≈1M BANs nightly without Kubernetes, via an app-level horizontal-scaling model launching one binary as N labeled worker instances, each claiming a DB-partitioned slice of work with runtime-tunable concurrency pools; raised throughput ≈10x (≈50 → ≈500 BANs/sec) with adaptive throttling.
- Engineered a Node.js/TypeScript (Fastify) multi-agent backend (scheduler, processor, web-service agents) running cron-driven automations with a custom concurrency/queue engine (libuv / p-limit) enforcing complex execution rules.
- Led the architecture and set the standards, boundaries, and runtime that every feature team built on, via a GraphQL Federation supergraph/subgraph platform (Node.js, TypeScript, Fastify, Mercurius).
- Built the core integration SDK: a singleton API connector unifying BSS and Legacy integrations with centralized auth, caching, timeout/retry, URL & API-version rewriting, and a normalized error taxonomy with source attribution and correlation IDs.
- Designed a 'plug-and-play' subgraph framework where teams provide only resolvers/business logic while the platform auto-wires federation, depth-limiting, error handling, context, caching, and hooks, cutting new-service bootstrap from days to minutes.
- Implemented a Kubernetes-native gateway that auto-discovers subgraphs via the K8s API and dynamically composes the federated supergraph with health polling, with zero manual gateway config when adding services.
- Architected a 3-tier system: a vanilla JS + PHP front end executing commands on remote Unix servers over SSH2 (phpseclib) from the browser, backed by two Java/Spring Boot microservices, one resolving any VAPP/hostname into its full environment topology from an Oracle registry, and a config plugin parsing XML app configuration into a searchable UI.
- Implemented role-based access control, full audit logging of every executed command (user, session, client IP, target host), and a modular widget framework (databases, customers, caching, ESB, message/SIM/IBAN utilities).
- Led the architecture and digital delivery of the Telco Back Office application, designing a micro-app (micro-frontend) architecture on Harmony alongside a new Java microservices backend.
- Took on Docker Swarm from scratch and led the production deployment design and rollout, turning an initially complex, unfamiliar orchestration model into a working end-to-end production deployment.
- Relocated to the North (Nazareth) to lead and mentor a full Scrum team delivering digital self-service for telecom clients, owning day-to-day delivery and technical direction.
- Repeatedly deployed as a senior front-end (React/Redux) specialist to stabilize and accelerate at-risk 'crisis' programs across European carriers, including Vodafone Italy (VFIT), 3UK, and 3IR.
- Marked the move from backend into the digital/full-stack world, building customer-facing self-service features with React and Redux, backed by Java/JDBC services over REST.
- Developed and maintained backend services in Java/JDBC for a large-scale telecom order management system, delivering features and resolving production defects.
Volunteer & Community Impact
An elderly relative of mine ran a 30-person group that split a shared daily reading cycle into one part per member. Every time the group finished a round, all 30 parts had to be handed out again — by hand, over a WhatsApp thread. Messages got buried, two people took the same part, others were skipped, and nobody could see what was still open. I built them an app to fix exactly that, and gave it away.
- Replaced the WhatsApp thread with a purpose-built app: create a group, invite members from contacts, assign each member a part, watch progress live, and roll the assignments over automatically when a round completes.
- Built deliberately for users aged ~45–70: Arabic RTL-first, large type, high contrast, one clear action per screen, and no ads, no growth mechanics — a sadaqa project, not a startup.
Built and deployed a production case-management platform in 4 days end-to-end during the October 2023 emergency, helping a municipality locate and support at-risk and displaced youth evacuated across the country.
- Centralized thousands of youth records into a single system used by counselors, program supervisors, and education leadership.