Ahmad BukhariBook a 25 min call ↗Agentic AI & LLM Systems / Resilient automation / CRM operations
Agentic AI, LLM systems and resilient automation, designed around how work actually breaks. This workflow is running now. Go ahead and break it.
Book a 25 min systems call ↗See the systems ↓
Verified n8n creator ↗github.com/syedahmad0786 ↗Available worldwide · Islamabad, Pakistan
The tools I wire together, stack-agnostic and failure-aware: OpenAI, Claude, n8n, Make, HubSpot, GoHighLevel, Vapi, ElevenLabs, monday.com, Power BI, Tableau, Airtable, Supabase, Intercom, Close, Next.js, Vercel, Python, FastAPI, Ollama, DuckDB, Pydantic, GitHub, Cal.com.
Client work · delivered through AiXCEL Solutions
Agencies, clinics, communities and investors handed over their lead flow, onboarding and reporting, and kept the systems we built. Counts describe the size of the work, not business results.
Built and handed over · AI agency · US
Problem. Lead, onboarding and reporting work had to run across four separate client workspaces.
What we built. Workflows across all four workspaces: ads, attribution, CRM and reporting.
Engagement completed · Smart vending · Community
Problem. Automations were spread across three platforms, and setting up a new member touched several tools.
What we built. Monitoring across every automation and one onboarding pipeline for new members.
Reconciled July 2026 · Healthcare marketing
Problem. A large set of GoHighLevel locations had to be accounted for before clinic accounts could run from one system.
What we built. An inventory of every location and one role-gated system for accounts and staff.
git log --graph · 7+ years operating
I spent years inside sales rooms, client conversations, training programs, and operational handoffs before designing the systems beneath them.
“That history shapes the work: I care whether the exception reaches the right person, whether the interface is clear at 2 a.m., and whether the team can operate the system after the builder leaves.”
Now
Architecting resilient CRM, voice, agentic, migration, and delivery systems for service businesses and internal teams.
2021 to 2024
Led teams, designed operating rhythms, and trained more than one hundred candidates across real estate organizations.
107 people trained across a 16-month leadership role · self-reported
2020 to now
Directing food and ration support: a human system where reliability is measured in families reached, not dashboards shipped.
748 ration packs mobilized in the COVID-response food program · self-reported
2019 to 2021
Translated complex property decisions into clear client choices across residential and commercial transactions.
AiXCEL Solutions ↗ · Founder
AiXCEL builds lead follow-up, client onboarding, reporting and monitoring inside each client’s own accounts, with a person approving every client-facing step. Founder-led, and built to be handed over with a runbook.
Four services · one approval layer · one set of numbers
01 · Subscription
Every enquiry captured, scored, routed to an owner and followed up on a clock until a call is booked.
02 · Sprint
From signed contract to live client without the chase. Anything that touches the client waits for approval.
03 · Subscription or Sprint
One agreed view of spend, leads and revenue, and a pulse in Slack each morning.
04 · Subscription
Every automation watched by deterministic checks. Failures arrive with the cause and a fix to approve.
Product
Business-owned AI with company memory, connected tools and role-based assistants. First deployed in my own agency, instead of renting a system I couldn’t inspect.
Product
AI search visibility audits. SIGNAL records what AI answer engines say about a business, which sources they cite, and whether the visits they send become qualified enquiries.
Modules inside the services: Onboard ↗ · ErrorLens ↗ · Pulse ↗ · AI Model Router ↗
Ten flagship agentic systems · verified public demos
A production-minded revenue and creator operations portfolio proving controlled multi-agent orchestration, evidence, typed APIs, human approval, evaluation, observability, replay, and cloud deployment.
Multi-agent campaign planning · Verified 5 Aug 2026
Five bounded roles turn a campaign objective and budget into an evidence-linked creator plan, then stop before outreach or spend.
LangGraph + FastAPI · 5 bounded roles · 3 scenario shapes · Objective-sensitive ranking · Approval gate before outreach or spend
Cross-system revenue assurance · Verified 10 Aug 2026
Ten bounded specialists reconcile advertising delivery, CRM state, funnel movement, and collected cash, then propose a recovery plan without changing any external system.
12 of 12 golden scenarios · 31 automated tests · 81.93% measured coverage · 25 Postman assertions · Signed serverless approval receipts · Docker and PostgreSQL acceptance
B2B forecast control · Verified
Twelve bounded stages compare seller confidence with exact buyer language, stakeholder coverage, and dated commitments, then stop at a sales manager decision.
12 of 12 golden deal states · 53 automated tests · 83.97% measured coverage · 28 live Postman assertions · PostgreSQL checkpoint restart proof · 0 external mutations
Creative measurement and learning · Verified
Eleven bounded stages separate attributable creative signal from sample, audience, placement, funnel, lag, fatigue, and data-quality confounds, then stop before publishing or changing spend.
13 of 13 golden scenarios · 68 automated tests · 82.76% measured coverage · 18 evaluation measures · 34 production Postman assertions · 0 platform mutations
Multilingual content operations · Verified 5 Aug 2026
Timed English scripts become distinct Urdu, Roman Urdu, or Arabic review packages with voice register, safety checks, and native-language approval.
FastAPI · 3 source scenarios · 3 locale routes · 3 registers · 27 meaningful combinations · Native-language approval
Independent evaluation · Verified 5 Aug 2026
A live black-box evaluator tests deployed systems for contracts, evidence, approval gates, boundaries, idempotency, and latency.
3 target systems · 7 weighted checks · 4 baseline and fault scenarios · Arbitrary URLs blocked
Pre-publish decision support · Verified 5 Aug 2026
A reproducible historical baseline returns forecast ranges, confidence, cohort fallback, and input sensitivity before publication.
500 licensed records · 400 training rows · 100 holdout rows · Versioned ridge models · No LLM claim
Governed revenue operations · Verified 10 Aug 2026
Seven bounded agents convert fragmented account evidence into an explainable qualification, speed-to-lead decision, and human-reviewed action proposal.
LangGraph + FastAPI, PostgreSQL path · 12 of 12 golden scenarios · 31 automated tests · 85.30% measured coverage · 18 Postman assertions · $0 replay inference
Creator and talent campaign control · Verified
Twelve bounded stages test audience authenticity, campaign fit, conflicts, rights, safety, historical evidence, budget, and concentration, then stop at a campaign manager decision.
13 of 13 golden scenarios · 70 automated tests · 85.38% measured coverage · 18 evaluation measures · 34 production Postman assertions · 0 external writes
Governed model access and release evidence · Verified 11 Aug 2026
A deterministic policy gateway gives nine deployed systems free replay, visitor-owned NVIDIA NIM or OpenRouter access, signed receipts, and an evidence-qualified evaluation registry without exposing an owner-funded model key.
Fixed provider allowlist · HMAC receipts · 85 automated tests · 18 of 18 evaluation measures · 40 production Postman assertions · 0 production 5xx
Open source · 6 public repos
Small, focused systems that expose the architecture, the safety rails, and the real run path.
Field notes · what I’m finding now
Next system · open to selected roles and builds
If the work crosses AI, automation, CRM, operations, or product delivery, and the edge cases matter, we should talk.
Book a systems call ↗Send the messy part ↗
Or ask an AI about me first: ChatGPT ↗ · Claude ↗ · Perplexity ↗ · Google AI Mode ↗ · Grok ↗ · Copilot ↗
Build in public · 8 channels
Watch the builds between the case studies.
Reels, walkthroughs, source and notes from systems in progress. Pick a channel.