I design, build, and hand off AI systems — automations, agents, RAG, cloud pipelines, and LLM features inside your product. Every engagement ends with docs, evals, and a team that can own it.
Client identities anonymised. Details available on request under NDA.
Every track, side by side.
Turn manual, repetitive work into reliable AI-driven workflows.
Ship AI automation systems that combine LLMs, APIs, and n8n workflows to remove manual work — from lead intake and content to reporting and internal ops.
Custom n8n workflows, nodes, and self-hosted deployments.
As an n8n Ambassador, I build production-grade workflows, custom nodes, and self-hosted setups for teams that need reliable automation without vendor lock-in.
Autonomous AI agents that plan, act, and complete tasks end-to-end.
LLM-powered agents that reason over tools, call APIs, and finish multi-step tasks without hand-holding — for support, ops, research, and content.
Google Cloud, BigQuery, and Vertex AI pipelines that ship.
Data pipelines and ML workflows on Google Cloud — BigQuery warehouses, Vertex AI training and serving, cost-efficient for teams beyond spreadsheets.
Redesign a broken process end-to-end, then automate it.
For ops stuck in email chains and spreadsheets — I redesign the process end-to-end, then automate what's worth automating.
Connect systems, ship webhooks, expose clean APIs.
REST APIs, webhooks, and integrations across the systems your business runs on — clean contracts, proper auth, observability from day one.
Ship AI features inside your SaaS — with evals, guardrails, and cost control.
Design and ship user-facing AI features — copilots, generators, semantic search — with eval harnesses, safety, and cost dashboards.
Grounded AI over your knowledge base — retrieval, reranking, evals.
Production RAG that retrieves the right chunks, grounds answers with citations, and stays accurate as your corpus grows.
Executive-ready dashboards backed by a modeled warehouse.
From raw data to decision-ready dashboards — modeled with dbt, served through Looker Studio, Metabase, or Power BI.
Voice agents on Vapi, Retell, and Twilio that answer, qualify, and book.
Inbound and outbound voice agents with CRM writes, calendar tool calls, warm human transfer, and full call analytics.
Fine-tune small open models to hit your quality target at lower cost and latency.
SFT and LoRA fine-tuning on Llama, Mistral, Qwen — or managed OpenAI/Anthropic tunes — for tasks where prompting hit its ceiling.
Golden datasets, evals, and CI gates so prompt changes ship without regressions.
The eval and observability layer your AI features need: golden datasets, rubric scoring, Braintrust/Langfuse, prompt registry, CI regression gates.
Package, serve, and monitor ML models on Vertex AI, Cloud Run, or Modal.
Production ML pipelines: training on Vertex AI, low-latency inference on Cloud Run or Modal, CI/CD, feature logging, drift alerts.
Kafka, Pub/Sub, and CDC pipelines that keep the warehouse in sync in seconds.
Event and CDC pipelines from product/marketing/IoT sources to BigQuery or Snowflake — Kafka, Pub/Sub, Debezium, materialized views.
dbt tests, lineage, and a catalog so every metric has an owner and a definition.
dbt tests, Great Expectations, column lineage, and a searchable catalog — so every metric has an owner, SLA, and shared definition.
Playwright bots for portals and admin panels without APIs.
Reliable Playwright/Puppeteer automations for legacy portals and admin panels — sessions, 2FA, layered selectors, LLM-assisted recovery.
LLM-assisted scrapers with proxies, dedup, and warehouse sinks.
LLM-assisted extractors that survive layout changes — managed proxies, fingerprinting, dedup, sinks into BigQuery/Postgres/CRM.
Chrome MV3 extensions with AI features, shipped to the Web Store.
Chrome/Edge MV3 extensions with LLM features — CRM enrichment on LinkedIn, AI drafts in Gmail, one-click admin actions — with OAuth and per-tenant config.
Fast, modern, SEO-ready websites built to convert.
Custom websites on React, Next.js, TanStack Start — server-rendered, accessible, Core Web Vitals optimized, shipped with CI/CD and a real CMS.
Shopify, WooCommerce, and headless storefronts that scale.
Custom e-commerce on Shopify, WooCommerce, or headless (Medusa) — fast PDPs, optimized checkout, product feeds, and every integration ops needs.
Multi-tenant web platforms with auth, billing, and dashboards.
Full-stack SaaS builds — auth, roles, billing, admin, dashboards — on TanStack Start/Next.js with Supabase or Postgres. MVP to production.
High-converting landing pages shipped in days, not months.
Focused landing pages and funnel flows for launches and lead gen — copy, design, analytics, and A/B testing wired in.
Ship a CMS your marketing team will actually use.
Headless CMS (Sanity, Contentful, Payload, Strapi) wired to a fast frontend — editorial workflows, previews, roles, and localization.
Ops team spent 20+ hours/week qualifying, enriching, and routing 400+ weekly inbound leads across email, forms, and Intercom.
"We reclaimed a full FTE of work and our SDRs now only see leads worth calling. Islam shipped it in 4 weeks."
Names and companies withheld — every quote is from a real client engagement, shared with role and industry only.
"The lead-routing automation shipped in 4 weeks. We reclaimed a full FTE of manual work and our SDRs now only see leads worth calling."
"Our Zapier bill was $4k/month and hitting task limits. 40+ workflows migrated to self-hosted n8n cut total cost of ownership by 72% with zero downtime."
"The support agent resolves ~65% of tier-1 tickets with cited answers. Response SLA went from 8 hours to under 3 minutes."
"Our BigQuery + dbt setup finally gave us one source of truth. Weekly reports build themselves and blended ROAS is visible in real time."
"Client onboarding went from a 5-day tangle to 45 minutes of async setup. It's the highest-leverage thing we've done for margins this year."
"A golden eval + CI gate on our drafting feature dropped hallucinated citations from 22% to under 2%. User trust transformed."
30-min call to map the problem, constraints, and success metric.
Fixed-scope plan with timeline, deliverables, and price.
Weekly demos, Loom updates, always-shippable increments.
Docs, runbook, training — your team owns it end-to-end.
Send a paragraph describing the problem via the form on this page or by email. I reply within one business day with either a scope-call invite or a referral if it isn't a fit.
A 30-minute scope call, a fixed-scope proposal with milestones and price, then weekly Loom updates and demos during the build. Every engagement ends with docs, a runbook, and a training session.
Every engagement fits one of three shapes: Sprint (1–3 weeks, fixed-scope), Build (3–8 weeks, fixed-price per milestone), or Retainer (monthly, ongoing). Exact numbers come out of a 30-minute scope call so the price actually matches your problem.
Small engagements (1–2 weeks of work) typically start within 5 business days. Larger builds go through a 2–3 week scoping phase first.
Yes — I ship PRs into your repo, pair with your engineers, and follow your review process. I can also run standalone if you don't have an in-house team yet.
SaaS, fintech, e-commerce, legaltech, healthcare, media, and agencies. If your problem shape is 'we have data or workflows that a human shouldn't be doing', I can probably help.
Data stays in your environment where possible. Secrets live in a managed vault, every workflow is audit-logged, and PII redaction is on by default for LLM steps. I'm happy to sign an NDA before the scope call.
Every AI engagement includes a golden dataset, eval suite, and a CI gate before any prompt or model change ships. That's how we catch regressions before your users do.
Yes — optional monthly retainers cover monitoring, on-call for critical workflows, and iterative improvements as your process evolves.
That's what the scope call is for. Describe the outcome you want, and I'll tell you honestly whether it's an automation problem, a data problem, or an AI-product problem — and what the right first step looks like.
A paragraph is plenty. I reply within one business day with either a scope-call invite or a referral if it isn't a fit.