Own your time
without the stress

Engineer who ships. I build and deploy production web platforms and AI products end-to-end — schema to UI to cloud.

End-to-end engineering.See availability

01 — About

The engineer behind the work.

I build and deploy production web platforms and AI products end-to-end. Schema to UI to cloud — no handoffs, no buzzwords.

Say helloSee the work
Portrait of Ahmad Shah, software engineer and tech founder

I bridge the gap between ambitious product vision and uncompromising systems engineering. From building high-throughput data engines like Kiln to deploying resilient platforms for Swiss enterprises, I design, code, and ship complete systems — with surgical taste and zero bloat.

Working with
Global clients
Open for
Q3 2026
Languages
EN · DE · AR · BN
Approach
Stack agnostic
See the work Let's talk

AI & Agents

Multiagent systems that plan, decide, and act across workflows — with the orchestration and governance production demands.

  • Agentic AI workflows
  • Multiagent architecture
  • Agent orchestration
  • Agent governance & security
  • LLM integration
  • RAG systems

Data & Cloud

AI-ready data platforms that agents can actually read — retrieval indexes, vector storage, agent-readable governance.

  • AI-ready data pipelines
  • Data lakes & warehouses
  • Data governance
  • Real-time streaming
  • ETL / ELT
  • Analytics platforms

Product Engineering

AI-native products and the agentic platforms around them — modern web, mobile, APIs, and agent operation centers.

  • Web & mobile applications
  • Cloud migration & architecture
  • API design
  • Kubernetes & containers
  • CI/CD pipelines
  • Legacy modernization

Stack I ship in production

PythonFastAPITypeScriptNext.js 16React 19PostgreSQLpgvectorRedisNVIDIA NIMAWSTerraformDockerGoNestJSgRPCTailwind CSSShadcn/UIVercel AI SDKPrismaKubernetesPythonFastAPITypeScriptNext.js 16React 19PostgreSQLpgvectorRedisNVIDIA NIMAWSTerraformDockerGoNestJSgRPCTailwind CSSShadcn/UIVercel AI SDKPrismaKubernetes
ECS FargateRDSElastiCacheCloudWatchGitHub ActionsCaddySQLAlchemyJWTStripePrometheusGrafanaSocket.ioYjsReact-Three-FiberFramer MotionZustandTanStack QuerySpring BootJavaNextAuthECS FargateRDSElastiCacheCloudWatchGitHub ActionsCaddySQLAlchemyJWTStripePrometheusGrafanaSocket.ioYjsReact-Three-FiberFramer MotionZustandTanStack QuerySpring BootJavaNextAuth
LATEST SHIP · FLAGSHIP SAAS BUILD

Kiln — know what enrichment costs before you run it.

A high-throughput waterfall enrichment & AI research engine that cuts B2B list costs by up to 81%. Billed only for verified results — misses, provider timeouts, and cached repeats are 100% free.

Launch Kiln LiveArchitecture & Case Study
~4¢
Per Verified Email
vs 20¢–40¢ on Clay (4–8 credits)
81%
Direct Savings
Retained on every outbound list
12+
Provider Adapters
Hunter, Apollo, Prospeo, Findymail...
$49/mo
CRM Sync Floor
Included on Starter (vs $495 gate)

Live List Cost Forecaster

Simulate pricing based on real provider rates

Interactive
5,000 contacts
1,00025,00050,000
45%

Cold B2B lists typically land between 35% and 55% verified matches.

Yields 2,250 verified work emails. What they cost:

Clay
$450
$0.20 each · 4 credits
Kiln (Platform)
$83
$0.037 each · 3 credits
Kiln (BYO Keys)
$65
$0.029 each · vendor at cost
Your List Savings
$367 (82% saved)
Core billing invariant: Provably Enforced
billed = hit && verified && !cached && !sample;

If no provider has the contact, or if a provider API times out, or if an email guess fails SMTP handshake verification, you are charged $0.00.

Waterfall · Outbound Run
1 Charge Made

Tomás Alvarez

vanta.com · Head of SecOps

Work Email
Findymailno record found
free ($0)
Icypeasrate limited · skipped
free ($0)
Prospeo (Verified Hit)tomas@vanta.com
0.90¢ charged
4
Huntershort-circuited
not called
3 provider calls madeCharged once only
AI AGENT COLUMN · NATURAL LANGUAGE RESEARCH

Instruct an AI agent to research each company row, validated against a strict JSON schema:

Company: "Vercel"
Prompt: "Does this company sell to developers?"
Schema Result: { sells_to_devs: true, evidence: "Per-developer seat pricing", confidence: 0.98 }

Supported native adapters & sync targets

HunterApolloClearbitProspeoFindymailDropcontactIcypeasZoomInfoPeople Data LabsHubSpotSalesforceInstantlySmartleadSlack

03 — Selected Work

Proven systems, all live in production.

Commercial SaaS platforms, customer storefronts, and deep-tech architectures where performance and security are provable, not promised.

View All Projects in Catalog
kiln.zelvante.com
Production interface of Kiln — Waterfall Data Enrichment & AI Research Engine
Latest Flagship Ship · Live in production
Zelvante / Ahmad Shah

Kiln — Waterfall Data Enrichment & AI Research Engine

Problem

Legacy data platforms charge 20–40¢ per email, charge upfront on failed lookups, and gate basic CRM sync behind $495+/mo enterprise plans.

Solution

Chains 12+ provider adapters into a waterfall that stops at the first verified hit. Transparent pre-run estimation; misses and timeouts are $0. Drops verified email cost to ~4¢.

Next.js App RouterTypeScriptTailwind CSSFramer MotionPostgreSQLLLM Structured Outputs
kiln.zelvante.comCase study
beautylounge-khatera.ch
Production interface of Premium salon booking platform
Live in production for Swiss clients
Beauty Lounge Khatera

Premium salon booking platform

Problem

A Swiss beauty salon needed a booking platform that matched their premium brand — fast, multilingual (DE/EN), and able to handle stylist portfolios + reservation flow without double-booking.

Solution

Next.js 16 App Router with ISR for service pages, Tailwind CSS design system, Framer Motion micro-interactions. Appointment state machine in PostgreSQL with row-level locking. Deployed on Vercel edge.

Next.js 16Tailwind CSSFramer MotionVercelPostgreSQL
beautylounge-khatera.chCase study
migmiggastro.ch
Production interface of Multilingual restaurant storefront
Live in production for Swiss clients
MigMig Gastro

Multilingual restaurant storefront

Problem

A restaurant needed a storefront that handled online ordering, real-time menu updates, and a kitchen display integration — 70% of their customers order from phones.

Solution

Next.js with locale-based routing (DE/IT/EN/TR), optimistic UI for cart updates, Server-Sent Events for order status, image optimization for food photography. Mobile-first interaction patterns.

Next.jsTailwind CSSSSEi18n RoutingImage Optimization
migmiggastro.chCase study
rag_tenant_isolation.py
RLS Verified ✓
@tenant_isolated_pipeline
async def query_rag(prompt: str, tenant: OrgContext):
# Hardware acceleration via NVIDIA NIM
vector = await nim.embed(prompt, model="nv-embed-v1")
sql = "SELECT doc FROM chunks WHERE org_id = :org ORDER BY v <=> :vec"
# 100% data leakage prevention enforced at database layer
PostgreSQL RLS Active
NVIDIA NIM · p99 18ms
Isolation enforced at the SQL query layer, provable live
Multi-tenant RAG Platform

Secure retrieval-augmented generation

Problem

Enterprise clients needed a RAG pipeline where tenant data isolation is provable — not just application-layer filtering, but enforced at the SQL query layer before vector similarity search runs.

Solution

FastAPI + SQLAlchemy 2.0 async + PostgreSQL RLS policies scoped per-request. NVIDIA NIM (nv-embed-v1 + llama-3.1-70b) for embeddings + token streaming over SSE. Isolation enforced at the SQL query layer, provable live.

FastAPIpgvectorNVIDIA NIMPostgreSQL RLSSSE
Live demo + 90-sec videoCase study

01 — How I Work

Four steps. Zero surprises.

Every engagement follows the same disciplined path — from a 30-minute discovery call to a 30-day post-launch observation window. You always know what happens next.

Discover

30 min call

We sit down for a 30-minute call. I ask sharp questions, map the bottleneck, and decide together whether AI/RAG, FinOps, or full-stack is the right lever to pull.

  • Architecture audit
  • Bottleneck diagnosis
  • 90-day success criteria

Architect

1–2 days

Before writing code, I draft the system design — data model, isolation boundaries, scaling ceiling, cost envelope. You sign off on the blueprint before a single line ships.

  • System design doc
  • Cost envelope model
  • Risk register

Build

Per sprint

Senior-only execution in two-week sprint commitments. You get a demoable build every Friday, deployed to your staging environment with full observability wired in.

  • Two-week sprints
  • Friday demos
  • Prometheus + traces from day one

Operate

30 days

Production handover with runbooks, dashboards, and on-call guidance. I stay on for a 30-day observation window to catch regressions before they reach your users.

  • Runbooks + dashboards
  • 30-day observation window
  • Knowledge transfer session

01 — What I do

Infrastructure rigor. Product polish. Same engineer.

I can architect your multi-region Kubernetes cluster and ship your customer-facing storefront — without handing either off to someone else.

Infrastructure

RAG that doesn't leak

Most multi-tenant RAG setups filter tenants in Python after the vector search. That gap is where data leaks. I push the tenant boundary into the SQL query plan, before pgvector runs.

FastAPI + SQLAlchemy 2.0 async + PostgreSQL RLS + NVIDIA NIM (nv-embed-v1, llama-3.1-70b). Token streaming over SSE. Isolation enforced at the SQL query layer, provable live.

FastAPIpgvectorNVIDIA NIMRLS
Infrastructure

Cloud bills that don't scale with traffic

If your AWS bill grows linearly with users, you're paying for on-demand when you could be paying for spot. I move stateless workloads to spot and shift them across regions hourly.

Terraform-managed spot fleets, 2-minute drain grace, Caddy at the edge. Stateless workloads move to spot and shift across regions hourly. The connection pool — not CPU — is usually the real bottleneck.

AWSECS FargateSpot FleetTerraformCaddy
Product

Commercial platforms that convert

A medical practice, a restaurant, a salon — they don't care about Kubernetes. They care that the site loads in 0.8s, books appointments without double-booking, and looks premium on a phone.

Next.js 16 App Router + Tailwind CSS + Framer Motion. ISR for content, edge functions for booking, multilingual routing. Live in production for Swiss clients.

Next.js 16Tailwind CSSFramer MotionVercel
Infrastructure

Legacy code that stops costing you hires

Tech debt isn't a code problem — it's an onboarding problem. When new hires take weeks to ship their first PR, the debt is winning. I measure debt in time-to-first-PR, not lines.

jQuery/JSP → React + Spring Boot, feature-flagged strangler pattern, SonarQube complexity as the metric. Phased migration with parallel-run validation behind a feature flag gateway.

ReactSpring BootFeature FlagsSonarQube

03 — Open Source

Things I build in the open.

Distributed systems, multi-tenant infrastructure, Kubernetes tooling, real-time collaboration.

Buraq

Distributed Task Queue

Highly concurrent, resilient distributed task queue built with Go and Redis Streams. Handles async jobs, scales workers, and manages failures through automatic retries and a Dead-Letter Queue.

  • 13,356 tasks/sec publish throughput
  • Automatic retries + DLQ with one-call replay
  • Real-time SSE event stream via Redis Pub/Sub
13.3K/s
Throughput
582ms
p99
10K
Concurrent
Go 1.24+Redis StreamsPrometheusGrafana
Star on GitHub

TenantKit

Multi-Tenant SaaS Boilerplate

Secure, scalable starting point so you don't rebuild auth, billing, and tenancy. PostgreSQL RLS per request via AsyncLocalStorage, JWT rotation with reuse detection, Stripe billing, Terraform AWS.

  • PostgreSQL RLS via SET LOCAL ROLE
  • JWT rotation with reuse detection
  • Stripe + signed webhook verification
  • Terraform AWS (ECS, RDS, ElastiCache)
6+
Concerns
11
Stack
Actions
CI
NestJS 11Next.js 16PostgreSQL RLSStripeTerraform
Star on GitHub

Sentinel-Aura

K8s Geographic Arbitrage

Kubernetes Operator and 3D Dashboard that visualizes and executes geographic compute arbitrage in real-time. Tracks Spot prices across 5 regions; operators trigger 1-Click Migrate when margin is profitable.

  • Multi-cluster orchestration via client-go
  • Safety dry-run: blocks if egress > savings
  • WebGL 3D Earth with custom ShaderMaterial
5
Regions
Live
3D
1-Click
Migration
Go Operatork8s/client-goR3FWebGLTailwind v4
Star on GitHub

Omni-Sync

Real-Time Collaborative Markdown

High-performance, real-time collaborative markdown editor for SRE Incident Playbooks. Go backend with CRDTs (Yjs), Next.js 15 frontend with Tiptap/ProseMirror, presence cursors, live K8s status blocks.

  • Yjs CRDT binary persistence in Redis
  • Floating presence cursors via Awareness
  • Inline K8sStatusBlock with live sparkline
  • Cmd+K command palette
Real-time
Collab
Zero-loss
Persistence
y-websocket
Protocol
Go + WSYjs CRDTNext.js 15TiptapRedis AOF
Star on GitHub

04 — Writing

Notes from production.

Deep dives on RAG isolation, FinOps, and backend work that makes SaaS scale.

View all posts 
# rag_search.py
async def rag_search(q):
  emb = await nim.embed(q)
  sql = "SELECT ... WHERE"
  sql += " user_id = :uid"
  res = await db.fetch(sql,
                       uid=user.id)
  # no leakage ✓
AI / RAG·Jul 2026

Eliminating Cross-User Data Leakage in Multi-Tenant RAG Pipelines

Why most multi-tenant RAG implementations leak data, and how enforcing per-user isolation at the SQL query layer — before vector similarity searches — closes the gap completely. Deep dive into pgvector + PostgreSQL RLS with code samples from a production FastAPI backend.

12 min readRead

05 — Social Proof

Voices that shape the work.

“Ahmad built our salon platform end to end — booking, stylist portfolios, the whole thing. It loads fast, looks premium, and we've never had a double-booking since launch.”

Khatera·Owner · Beauty Lounge Khatera

05 — The Execution Loop

How I actually operate.

High-ticket clients care about transparency. Here's the exact delivery mechanism — no black boxes, no vague timelines.

01

Core Telemetry Audit

I clone your stack, map database boundaries, and hunt for immediate low-hanging infrastructure leaks before any contract is signed.

  • Stack + database boundary map
  • Infrastructure leak report
  • Quick-win cost cut list
02

Bi-Weekly Sprints

Direct Slack channel, PR updates every 48 hours, sandbox deployments. You see code moving twice a week — no black-box engagment.

  • Dedicated Slack channel
  • PRs every 48 hours
  • Friday sandbox demos
03

Clean Handover

Documented Terraform state, automated CI pipelines, and an onboarding session for your engineering team. You own everything I touch.

  • Terraform state files documented
  • CI pipelines automated
  • Team onboarding session
Q2 2026: 1 Strategic Partnership Slot Available

06 — Strategic Partnerships & Engagements

Architectural excellence. Guaranteed outcomes.

I don't bill by the hour or pad teams with junior middlemen. Choose an engagement track calibrated for velocity and measurable business ROI.

Schedule a 15-Min Scoping Call
1-Week Intensive

Architecture & FinOps Diagnostic

A surgical 5-day deep dive into your infrastructure, database bottlenecks, or runaway cloud costs. Uncovers hidden leaks and delivers actionable, benchmarked code diffs.

Core Focus

  • Codebase & DB query plan performance audit
  • Cloud spend (FinOps) analysis targeting 30–70% cost reduction
  • Multi-tenant data isolation & RAG security review

Tangible Deliverables

  • Executive architecture roadmap & risk matrix
  • Ready-to-merge PRs with performance benchmarks
  • 60-min recorded architectural walkthrough call
Ideal Fit: Founders & CTOs preparing for high-traffic scale, fundraising, or cloud bill crises.
Book 1-Week Diagnostic
Most Requested by Founders
Bi-Weekly Sprints

Embedded Principal / Fractional CTO

A senior technical partner embedded directly into your engineering team. Hands-on architectural design, production-critical shipping, and zero junior delegation.

Core Focus

  • Direct code commits to your core repositories
  • Zero-leakage RAG, AI agent columns, & microservices
  • CI/CD automation, cloud orchestration, & edge architecture

Tangible Deliverables

  • Bi-weekly production releases & working staging demos
  • High-signal code reviews & system design RFCs
  • Direct async communication via dedicated Slack/Discord
Ideal Fit: Fast-moving startups needing top-tier engineering velocity without a $400K+ full-time executive hire.
Check Sprint Availability
Fixed Milestone

Turnkey Production Platform Handover

Complete ownership of an MVP, commercial storefront, or enterprise platform from system architecture and pixel-perfect UI to production edge deployment.

Core Focus

  • End-to-end full-stack build (Next.js 16 + PostgreSQL)
  • High-converting UX/UI with tactile micro-interactions
  • Row-level security, payment gateways, & i18n localization

Tangible Deliverables

  • 100% intellectual property & clean codebase handover
  • Automated testing suite & 95+ Lighthouse score guarantee
  • 30-day post-launch hyper-care & monitoring support
Ideal Fit: Commercial enterprises, Swiss luxury brands, or founders launching a new flagship product.
Request Platform Proposal

Built on 4 Non-Negotiable Guarantees

Eliminating contractor ambiguity with concrete engineering standards.

100% IP & Codebase Ownership

Every line of code, test, and infrastructure script belongs entirely to your company from day one. Zero proprietary vendor lock-in.

Direct Principal Execution

Zero junior bait-and-switch. Every architectural decision, database migration, and pull request is authored directly by Ahmad Shah.

Provable SLAs & Benchmarks

Real metrics backed by mathematical guarantees: 95+ Lighthouse scores, sub-300ms p99 latencies, and PostgreSQL RLS tenant safety.

Working Software Every 2 Weeks

No months of silence. Every sprint yields testable deployments on live staging URLs, letting you validate progress in real time.

Interactive Scoping Assistant

Estimate your project fit & timeline

Select your primary engineering objective below to view the recommended engagement track and sprint structure.

1. What are you looking to architect or solve?

2. Target Kickoff Window

Recommended Model

Embedded Principal (2-Week Sprints)

Estimated Timeframe
2 to 4 Sprints

FastAPI + pgvector + PostgreSQL RLS + NVIDIA NIM embedding pipeline with provable data isolation.

Schedule Scoping Call (15-min)Send Brief via Email

The future of SaaS is AI-native

Let's build it together.

Whether you need a RAG pipeline that respects tenant boundaries, a cloud bill that doesn't grow with traffic, or a full SaaS shipped on a deadline — I'm taking on one new engagement in Q3 2026.

Schedule a Strategic ConsultationSend a message

07 — Contact

Tell me what you're building.

Whether you need a RAG pipeline that respects tenant boundaries, a cloud bill that doesn't grow with traffic, or a full SaaS shipped on a deadline — let's scope it in a 30-minute call.

contact@ahmadstx.com

Typical reply: under one business day.

Command Palette

Search for a command to run...