Jafer Mohammed.
AI Product & Platform Builder

Jafer Mohammed

AI Product Lead · Forward Deployed AI Engineer · AI Solutions Architect
Founder, AqionLabs.ai (AqionLabs FZ-LLC, RAKEZ) · Dubai, UAE · English, Malayalam, Hindi
9 yrsbuilding & operating production systems
4 yrshands-on AI / GenAI as a forward deployed engineer
6 yrson banking-sector projects
11live AI prototypes, products & web apps
01

Summary

Who I am, mapped to AI platform roles

I build AI products end to end: from customer discovery and roadmap, through architecture and hands-on engineering, to production hand-over. I have nine years of experience building and operating production systems (six in software development and six in technical project management, overlapping), including four years hands-on in AI/GenAI. At AqionLabs.ai, the AI company I founded in Dubai, I own the roadmap, solution architecture and go-to-market, and I personally build the agentic, voice and retrieval systems, including the live EC Calling Agent pilot and the Business Brain for EthikCorp.

Multi-model & agentic systems

  • Multi-agent orchestration (LangGraph), MCP servers, and FastAPI tool-calling with schema-validated LLM arguments
  • Fine-tuning plus RAG to build a client's own company model
  • Model routing, semantic caching, token optimisation and inference-cost management; LLM vendor evaluation (AWS Bedrock, Azure AI Foundry)

Data acquisition & pipelines

  • n8n pipeline: Apollo sourcing → Firecrawl/Apify enrichment → LLM-personalised copy → send
  • Knowledge pipelines: inventory, cleaning, de-duplication, PII redaction, chunking with metadata, embeddings, hybrid retrieval and reranking
  • Census-cleaning and website-ingestion prototypes

Compliance, grounding & evaluation

  • PII redaction, data governance, UAT and regression suites on a live client agent; UAE PDPL-aware data handling
  • Groundedness and task-success evals (pytest, DeepEval) as a CI regression gate
  • Sentence-level evidence links, human approval steps, role-based retrieval and tamper-evident / hash-chained audit trails

Enterprise delivery & product ownership

  • Delivery for healthcare, banking and global enterprise customers (incl. Bank of America, City National Bank, Christus Health, McKinsey, Allstate)
  • Led cross-functional squads without formal authority, from discovery through SIT/UAT and cutover to hand-over
  • A portfolio of AI products and prototypes on shared Cloudflare infrastructure under AqionLabs
02

Achievements

Oct 2026
Google Immersion EMEA — Phase 1

Selected for Google Immersion EMEA Phase 1, powered by Antler.

Meta Wearables Studio UAE

Selected for Meta Wearables Studio UAE.

ElevenLabs grant — top 5 of ~10,000

Top 5 in the ElevenLabs × Ignyte Voice AI Challenge (~10,000 participants); received an ElevenLabs grant.

ICC × Ignyte Demo Day

Selected to present a live AI prototype at the ICC × Ignyte Global Hackathon Demo Day, Dubai AI Festival, 27 Oct 2026.

03

Projects & Prototypes

Client work, enterprise projects, live prototypes
Client pilotLive ↗ · GitHub

EC Calling Agent and Business Brain for EthikCorp, Dubai

A live voice AI calling agent for EthikCorp's sales and operations, with a company 'business brain' (Aqion Superbrain) behind it.

Role
Forward deployed lead: owned discovery, design, engineering, security, deployment and KPI measurement while embedded with EthikCorp's teams.
Tech
FastAPI tool-calling (lead capture, callback booking, escalation) with schema-validated LLM arguments · multilingual (English, Arabic, Hindi) agent workflows · Retell/Vapi · Supabase · fine-tuning + RAG · Cloudflare Workers + D1
Outcome
Taken from prototype to production with PII redaction, data governance, UAT, prompt A/B testing, regression suites and operational hand-over.

EthikCorp Business Brain (Aqion Superbrain)

EthikCorp's own company model: I fine-tuned a base model and grounded it with retrieval-augmented generation (RAG) over EthikCorp's company knowledge, so answers come with sources and only reach people allowed to see them.

  • Data readiness portal (Phase 1). A Cloudflare Worker + D1 portal takes EthikCorp through nine stages: inventory of 18 sources with named owners, collection, cleaning and de-duplication, PII handling under UAE PDPL, heading-based chunking with metadata (source, owner, department, sensitivity, version, hash), embeddings and a vector store for hybrid search, role-based access enforced inside retrieval, evaluation, and a refresh cadence.
  • RAG vs fine-tuning, decided per source. Facts, policies, SOPs, CRM records and contracts are retrieved through RAG, with restricted tiers for legal material and PII. Brand voice, proposal style, call scripts and report formats feed the fine-tuning set.
  • Evaluation design. A golden set checks citations, refusals, conflicting document versions and permission leaks, with a target of zero leaks.
  • Superbrain workspace. Department workspaces (Sales & BD, Marketing, Operations, Finance, HR & Talent, Legal & Compliance, Client Success, Consulting Delivery, Technology & AI) with reviewed output templates such as proposal briefs, weekly status briefs, minutes and action registers, and ROI evidence packs, plus a CEO view.

Status: in pilot with EthikCorp, not generally available. Both portals are access-controlled, so they are not linked here.

AqionLabs

Agentic AI and enterprise knowledge automation

Reusable multi-agent and RAG foundations behind AqionLabs client builds.

Role
Architect and hands-on builder.
Tech
LangGraph multi-agent orchestration · RAG (chunking, embeddings, pgvector, hybrid retrieval, reranking) · GitHub Actions CI/CD for agents and prompts · LangSmith/OpenTelemetry · pytest + DeepEval
Outcome
Automated LLM-eval regression gate for groundedness and task success. Builds turned into accelerators: a tool-calling service template, guardrails, an eval harness and reference architectures.
AqionLabs

AI outbound pipeline

End-to-end lead pipeline: Apollo lead sourcing → Firecrawl/Apify web enrichment → LLM-personalised copy → Gmail/Apollo dual send.

Role
Designed and built it with Claude and Codex.
Tech
n8n · Apollo · Firecrawl · Apify · LLMs · Gmail
Outcome
A running outbound workflow for AqionLabs GTM.
Work project

Patient communication modernisation and AI-ready contact centre — American Hospital Dubai

Redesigned IVR flows, intelligent routing and call recording for American Hospital Dubai and 8 clinics (2,500+ users).

Role
IT Voice Lead Engineer: led telecom, network, CRM, vendor and business teams without direct authority.
Tech
Avaya Aura (AEP, AACC, CM, SM, SBC, POM, ACR) · CRM workflows · Grafana
Outcome
Integration and data foundation for later conversational AI, speech analytics and voice-agent use cases; simpler KPI reporting with Grafana dashboards.
Work project

Gateway infrastructure refresh — Christus Health, USA (Servion)

Global refresh replacing 100+ G650 gateways with 300+ G450 gateways and media servers.

Role
Senior Consultant: led an implementation squad of architects, PMs, QA, vendors and client teams.
Tech
Avaya Aura · Python validation scripts · MOPs and cutover plans
Outcome
Better scalability and resilience; Python pre/post-cutover validation cut manual cutover errors.
Work project

Applications on Avaya OneCloud Private — McKinsey, UIHC, Allstate (Betsol)

Cloud-native platform deployments for enterprise customers.

Role
Cloud Implementation Engineer.
Tech
Docker · Kubernetes · VMware ESXi (Python automation) · Swagger/OpenAPI · APIGEE · Jenkins CI/CD · Palo Alto
Outcome
Standardised deployments with Avaya Reference Architecture teams; improved observability and security.
Work project

Avaya servers, NAR region — Apple, CVS, UPS, Progressive (Innovatia)

Tier 3 support for business-critical communications infrastructure.

Role
Technical Support Engineer.
Tech
Avaya / Cisco UC · MST, traceSM, traceSBC · Python log parsing
Outcome
24/7 service continuity; Python log parsing to find recurring failure patterns.
PrototypeLive ↗ · GitHub

SANAD — evidence chain for corporate banking

Turns a governed evidence set into a client-review and credit-review brief. Built for the Ignyte × Warba Bank challenge (synthetic data).

Role
Solo builder.
Tech
Sentence-level evidence links · deterministic field validation · human review routing · tamper-evident audit pack · Cloudflare Pages
Outcome
Live MVP plus a full governance pack (threat model, AI governance, Shariah governance notes).
PrototypeLive ↗ · GitHub

Mizan AI — policy war-room

Multi-agent incident simulator for UAE digital-government AI incidents. ITU AI for Good Hackathon UAE 2026 entry (synthetic data).

Role
Solo entrant.
Tech
Multi-agent · cited UAE and global policy sources · hash-chained audit · Python · Workers + GitHub Pages
Outcome
Live replay, a 4-page technical report and a portable knowledge-base export.
PrototypeLive ↗ · GitHub

Aqion Relay — ResilienceTech MVP

Business-resilience workbench: explore synthetic supply-recovery options while protecting operating cash. du × Ignyte concept.

Role
Solo builder.
Tech
Next.js · TypeScript · Vitest · Cloudflare Workers
Outcome
Live MVP with a pitch deck and a system-design document.
PrototypeLive ↗ · GitHub

Journey Companion

Employee-side AI companion for onboarding and internal mobility. Emirates Group × Ignyte entry (synthetic data).

Role
Solo builder.
Tech
Cloudflare Workers AI · JavaScript · e2e tests
Outcome
Live prototype, a captioned demo video and a 2-page solution note.
PrototypeLive ↗

Boundary — women's cricket editorial desk

Finds under-covered women's cricket performances and generates source-linked story drafts, with a separate editor-approval step. ICC × Ignyte concept (demo data).

Role
Builder.
Tech
Source-linked generation · human approval step
Outcome
Live public demo.
PrototypeLive ↗

Quote in Seconds — SME group medical quoting

Census-to-quote automation that cleans messy census files with confidence-scored fixes, re-rates live and triages underwriting. Cigna × Ignyte prototype (fictional data).

Role
Builder.
Tech
Data-cleaning pipeline · OpenAPI 3.1 · Cloudflare Workers
Outcome
Live prototype.
PrototypeGitHub

Human-gated pre-authorisation agent

Provider pre-authorisation workflow for the Ignyte × ElevenLabs challenge (synthetic data).

Role
Builder.
Tech
Python · consent gates · strict schemas · HMAC webhook verification · append-only audit · 30-case eval suite
Outcome
The agent has no final-decision method; only a separate human approver route can record an outcome. Runs locally.
ProductLive ↗

JobCaptain

A truthful AI job agent for the UAE: it scans employers' career sites on a schedule, checks every requirement against the real CV and prepares tailored material.

Role
Founder and builder.
Tech
Agentic workflow · scheduling · Cloudflare Workers
Outcome
Live landing page.
ProductGitHub

AqionVoiceAIBuilder

Builds a client-branded, multilingual voice-agent demo from a client's website URL, with a built-in CRM and call transcripts.

Role
Founder and builder.
Tech
React · Express · Retell · Firecrawl (with a safe public-page fallback)
Outcome
Reusable studio for AqionVox client demos (needs server keys, so there is no public demo).
ProductLive ↗ · GitHub

AqionVox MVP dashboard

Conversational-AI platform dashboard: leads, conversations, call logs, meetings, WhatsApp, voice-AI cost and conversion funnel (mock data).

Role
Builder.
Tech
React · Vite · TypeScript · Recharts
Outcome
Live demo.
ProductLive ↗ · GitHub

Aqion Voice operations dashboard

Voice-AI operations dashboard: call transcripts, leads, bookings and voice personas (demo data).

Role
Builder.
Tech
React · Vite · TypeScript · Supabase-ready
Outcome
Live demo.
CompanyLive ↗

AqionLabs.ai

Company site: agentic AI that answers calls, qualifies leads and moves work into business systems.

Role
Founder.
Tech
React · Vite · Tailwind
Outcome
Live.

Links at a glance

ProjectPublic URLGitHub
EC Calling Agent and Business Brain for EthikCorp, Dubaiethikcorp.aqionlabs.comECcallingAgent
SANAD — evidence chain for corporate bankingsanadaqionlabs.pages.devSanad_IgnyteWarbaBank_Challenge
Mizan AI — policy war-roommizan-ai-itu-uae.aqionlabs.workers.devmizan-ai-itu-uae
Aqion Relay — ResilienceTech MVPduignytemvp.aqionlabs.workers.devDuIgnyteMVP_Aqionlabs
Journey Companionjourney-companion.aqionlabs.workers.devemployee-journey-companion
Boundary — women's cricket editorial deskboundary.aqionlabs.com/web—
Quote in Seconds — SME group medical quotingcigna-quote-in-seconds.aqionlabs.workers.dev—
Human-gated pre-authorisation agent—ELEVENIGNITECHALLENGE
JobCaptainjobcaptain.aqionlabs.workers.dev—
AqionVoiceAIBuilder—AqionVoiceAIBuilder
AqionVox MVP dashboardaqionvox-mvp-aqionlabs.aqionlabs.workers.devaqionvox_mvp
Aqion Voice operations dashboardaqionvoicemvp-aqionlabs.aqionlabs.workers.devAqionVoiceMVP
AqionLabs.aiaqionlabs.ai—

Challenge prototypes use synthetic or fictional data, as each one states on its own page.

04

Experience

2017 – present

AI Product Lead

AqionLabs.ai, UAE (founder)
Jan 2026 – Present

EC Calling Agent and Business Brain for EthikCorp

  • Embedded with EthikCorp's sales and operations teams; owned the full journey from business problem through design, engineering, security, deployment and KPI measurement.
  • Built the agent's tool-calling with FastAPI: lead-capture, callback-booking and escalation endpoints with schema-validated LLM-generated arguments; English, Arabic and Hindi agent workflows with multi-step orchestration.
  • Took the agent from prototype to production, owning PII redaction, data governance, UAT, prompt A/B testing, regression suites and operational hand-over.
  • Built EthikCorp's Business Brain by fine-tuning a base model and grounding it with RAG over a governed, permission-aware knowledge pipeline.

Agentic AI and enterprise knowledge automation

  • Designed multi-agent orchestration (LangGraph) and RAG pipelines: chunking, embeddings, vector search (pgvector), hybrid retrieval and reranking.
  • Set up MLOps/LLMOps foundations: CI/CD for agents and prompts (GitHub Actions), prompt/model versioning, observability and an automated LLM-eval regression gate; built evaluation harnesses (pytest, DeepEval) for groundedness and task success.

Additional responsibilities

  • Built an n8n outbound workflow (Apollo lead sourcing, Firecrawl/Apify enrichment, LLM-personalised copy, Gmail/Apollo dual send) using Claude and Codex.
  • Turned client builds into reusable accelerators (tool-calling service template, guardrails, eval harness, reference architectures) and a documented forward-deployed delivery methodology.
  • Owned AI roadmaps, customer discovery, solution architecture and GTM; evaluated LLM vendors and cloud AI services (AWS Bedrock, Azure AI Foundry) on accuracy, latency and security.

IT Voice Lead Engineer (AI and Enterprise Communications)

American Hospital, Dubai
Apr 2024 – Sep 2025
  • Embedded with hospital operations and contact-centre teams to redesign IVR flows, intelligent routing and call recording for American Hospital Dubai and 8 clinics (2,500+ users).
  • Built the integration and data foundation (AEP, AACC, CRM workflows) for later conversational AI, speech analytics and voice-agent use cases.
  • Led telecom, network, CRM, vendor and business teams without direct authority through implementation and operational hand-over; simplified KPI reporting with custom Grafana dashboards.
  • Managed Avaya Aura SMGR, CM, SBC, SM, AACC, AEP, POM and ACR, keeping production reliable through monitoring, incident resolution and vendor escalation; wrote runbooks and SOPs.

Senior Consultant (Enterprise Voice)

Servion Global Solutions, Bangalore
Nov 2022 – Apr 2024
  • Delivered the Christus Health (USA) global refresh: 100+ G650 gateways replaced with 300+ G450 gateways and media servers; Python pre/post-cutover validation scripts reduced manual cutover errors.
  • Led an implementation squad (architects, PMs, QA, vendors, client teams) without formal authority using MOPs, cutover plans and validation checklists.
  • Designed scalable Avaya Aura UC/CC architectures for automation, intelligent routing, IVR modernisation and speech workflows; implemented API integrations, SIP trunking and SIPREC media forking for real-time speech/AI analytics.
  • Ran technical discovery, solution workshops, SIT/UAT, cutover and production hand-over for strategic customers including Tesla, UPS, Disney, Bank of America, UIHC, Christus Health, TeleTech and City National Bank; supported production stability through root-cause analysis and escalation management.

Cloud Implementation Engineer

Betsol, Bangalore
Sep 2021 – Nov 2022
  • Designed and deployed cloud-native platforms on Avaya OneCloud Private, Docker, Kubernetes and VMware for McKinsey, UIHC and Allstate; automated VMware ESXi provisioning and VM cloning with Python.
  • Built orchestration workflows with Swagger/OpenAPI, Postman and the APIGEE gateway; ran Jenkins CI/CD with version-controlled artifacts and automated validation.
  • Resolved third-party integration issues with Avaya Reference Architecture teams; improved observability and security (monitoring, traffic analysis, Palo Alto Firewall/Panorama, access control).

Avaya and Cisco Technical Support Engineer

Innovatia India Pvt Ltd, Bangalore
Nov 2017 – Sep 2021
  • Tier 3 support for business-critical communications infrastructure, maintaining 24/7 service continuity for Apple, CVS, UPS and Progressive.
  • Investigated complex faults with MST, list trace, traceSM and traceSBC; wrote Python log-parsing scripts to find recurring failure patterns.
  • Led Cisco UC implementation and configuration for high availability; partnered with product engineering on critical patches and managed escalations and root-cause investigations.
05

Education & Certifications

B.Tech CS · 11 certifications

Bachelor of Technology in Computer Science

MES College of Engineering, University of Calicut, Kerala · 2011 – 2015

Google Certified AI ProfessionalGoogle
Microsoft Certified: Azure FundamentalsMicrosoft
ITIL V4Axelos
Certified Scrum MasterScrum Alliance
Oracle Database 11g Administrator Certified AssociateOracle
Avaya Certified Implementation SpecialistACIS 7120
Avaya Experience Portal Administrator SpecializedASAC-0032
Avaya Aura Communication Manager AdministratorASAC-0038
Avaya Aura Call Centre Elite AdministratorASAC-0018
Avaya Aura Session Border Controller AdministratorASAC-0039
Avaya Aura System Manager AdministratorASAC-0037
PMP — in progressPDUs earned; certification not yet completed
06

GitHub Repos

Open-source candidates

Selected original repositories at github.com/Jaferdecodesai with reusable value beyond a single client, and the clean-up planned before each is formally open-sourced.

RepositoryWhy it is worth open-sourcingBefore release
mizan-ai-itu-uaeMulti-agent policy war-room with cited sources and a hash-chained audit; already MIT-licensed with a NOTICE file.Confirm third-party licences under third_party/ (ITUAIReadiness), and that the KB chunks only quote sources that allow redistribution.
Sanad_IgnyteWarbaBank_ChallengeReference design for evidence-linked, human-reviewed document AI in banking, with a governance and threat-model pack.Add a LICENSE file (only a LICENSE_INVENTORY exists); drop the D1 database_id from wrangler.jsonc or make it a placeholder.
ELEVENIGNITECHALLENGEDependency-free human-in-the-loop agent pattern: consent gates, an approver-only decision route, HMAC webhooks and a 30-case eval suite.Add a LICENSE; keep the demo approver token clearly demo-only; move TEAM.md contact details out if needed.
employee-journey-companionSmall, readable Cloudflare Workers AI companion app with tests and e2e scripts.Add a LICENSE; consider neutral naming and remove Emirates Group challenge branding from the UI.
DuIgnyteMVP_AqionlabsNext.js-on-Workers MVP template with tests, a Cloudflare build script and architecture docs.Add a LICENSE; strip deployment evidence files (worker version IDs, account details) and the unused mysql2 dependency.
AqionVoiceAIBuilderReusable 'voice-agent demo from a URL' studio: website ingestion, knowledge files, Retell calls and a browser-side CRM.Add a LICENSE; keep secrets in .env only (an .env.example exists); review the server for hard-coded SMTP or Retell values before publishing.
aqionvox_mvpClean React + Recharts dashboard template for conversational-AI products (mock data only).Add a README, a LICENSE and screenshots.