I ship AI-assisted products with proof, control, and fallback logic.
I architect the financial layer founders can actually audit.
I turn ambiguous requirements into shipped, maintainable systems.
Full-stack engineer specializing in AI-assisted products, transactional correctness, financial infrastructure, and startup architecture. I build systems that move real money, preserve state under failure, survive scaling pressure, and stay maintainable under startup velocity.
Toptal accepts fewer than 3% of applicants. Available for consulting engagements, technical advisory work, and high-ownership startup engineering through Toptal or directly.
Available for work
Open to roles worth doing.
Consulting, advisory, and high-ownership engineering. Especially where the system needs to be correct, not just functional — AI products, fintech, SaaS, and startups approaching their next inflection point. Remote worldwide.
Not a list of technologies. Six capability areas that map directly to the problems startups face when they grow past their first system.
Financial Infrastructure
Double-entry ledgers, wallet systems, cent-accurate reconciliation, idempotent payment workflows. I build the financial layer that product teams can reason about, operators can audit, and founders can sleep through.
Double-entryStripePlaidReconciliationBNPLLedgers
Transactional Systems
Idempotent workflows, retry pipelines, event-sourced state, queue-based processing, and deterministic state transitions. Systems that stay correct when webhooks retry, jobs fail twice, and three requests land at once.
BullMQIdempotencyEvent-drivenRedisWebhooks
Startup Architecture
Monorepo structure, API-first design, shared packages, deployment pipelines, auth, admin dashboards, and product workflows. Architecture that lets two engineers move at the pace of ten — without accumulating the debt that kills momentum at Series A.
MonorepoOpenAPIFastifyPostgreSQLDocker
Platform Modernization
Legacy system evolution without risky rewrites. I stabilize what exists, surgically upgrade what matters, and replace only what cannot scale — without a rewrite that freezes the roadmap for six months.
RefactoringMigrationAWSCI/CDObservability
Data & Automation
Scraping pipelines, batch processing, ETL workflows, report generation, and long-running workers. Reliable data infrastructure that handles job failures gracefully, resumes where it left off, and doesn't wake you up at 3am when volume spikes.
Web scrapingETLBullMQCronCSV/Reports
AI-Assisted Products
Structured context systems, browser capture, embeddings pipelines, semantic clustering, and LLM-assisted workflows with deterministic fallbacks. AI that ships as product — with validation, deterministic fallbacks, and the operational discipline you'd apply to a payment workflow.
Each role mapped to the startup lifecycle phase it represents — because the challenges of 0→1 are completely different from 10→100.
0 → 1 · Founder product
Motiavated
Founder · Sole Full-stack Engineer
Private betaRemoteAI · Job Application Workspace
AI Productivity
What I built & why it matters
Built a private-beta job-application workspace that captures opportunities, maintains a structured Experience Vault, and generates evidence-backed application answers and resume JSON outputs without auto-applying.
Shipped browser-extension capture, server-side job parsing, Vault-grounded answer generation, onboarding, Google login, credit tracking, admin dashboards, public landing pages, and self-hosted infrastructure.
In private beta: 4 users, 80 captured opportunities, 235 generated application answers, and 10 Reactive Resume-compatible JSON outputs — all without auto-applying.
Designed the product around manual review and human control — better apply, not auto-apply — with a planned PDF path using self-hosted Reactive Resume, service-account proxy downloads, and later OIDC user accounts.
Architected a Yarn 4 monorepo with shared packages for auth, UI, security, and observability — giving a small team the structure of a larger engineering org.
Designed and built double-entry ledger and wallet systems with deposits, withdrawals, reserved balances, fees, and reconciliation-safe accounting — the financial backbone the platform runs on.
Implemented Stripe deposits, Plaid withdrawals, trading, IPO precommitment, NAV, and membership billing — complex financial workflows that require correctness, not just functionality.
Established AGENTS.md and repo-level AI documentation — making Claude Code more reliable with deterministic, auditable development behavior.
Architected a complete BNPL platform — backend, cross-platform mobile app, admin CRM, and marketing site — from zero in a monorepo.
Designed the full lending lifecycle — investor funding, plan creation, virtual card issuance via Stripe Issuing, installment scheduling, delinquency, recovery, and expiry — all auditable and cent-accurate.
Built immutable investor and platform ledgers with wallet accounting, repayment distribution, loan assignment/release, and profit-share calculations.
Implemented idempotent webhook ingestion and auditable workflow routing — making the system safe to retry under failure conditions.
Built an embeddings pipeline that vectorizes Shopify and Etsy product catalogs, stores vectors in PostgreSQL, and clusters products into semantic niches using cosine similarity.
Integrated an LLM naming layer with strict output validation and deterministic fallback — making AI-generated taxonomy stable and safe for business users to rely on.
Designed the system around explainability and stability, combining semantic clustering with deterministic fallback logic.
Upgraded deprecated Elastic Beanstalk infrastructure and modernized legacy packages — stabilizing a platform under scaling pressure without a risky full rewrite.
Optimized large-marker map performance using server-side and client-side caching with dynamic grouping by zoom level — solving a UX problem that had been blocking adoption.
Migrated mapping from Mapbox to Google Maps under deadline and improved scalability using queue-based workflows — two high-risk deliveries shipped on time.
Built resilient state-specific license scrapers with batching, retry/recovery, resume support, duplicate detection, and failure classification — turning fragile scripts into a reliable pipeline.
Modernized auth, team management, admin/editor roles, security settings, CI, Docker, and test workflows incrementally without breaking production.
Vetted in the top 3% of engineers on Toptal — a rigorous screening process covering technical depth, communication, and delivery reliability.
Delivered complete products across fintech, energy, logistics, AI productivity, and e-commerce — working directly with founders, CTOs, and product owners across ambiguous, fast-moving environments.
Node.jsTypeScriptReactPostgreSQLStripeAWSDocker
Selected work
Past Projects
01 / 04
Logistics · Marketplace
Dromigo
Two-sided logistics marketplace with GIS freight matching, real-time driver tracking, and Stripe marketplace payments. Co-founded and shipped from zero.
Node.jsFlutterPostGISStripeWebSocket
02 / 04
Fintech · Virtual Cards
Rassidi
Fintech PWA for virtual payment cards in local currency, with real-time inventory, push notifications, API-driven workflows, and a mobile-first experience.
Nuxt.jsPWAAPIOneSignal
03 / 04
B2B SaaS · CRM
Adgon Expert
Full CRM for expert geometers with a public-facing professional search engine, subscription billing, and tender/news feeds per plan tier.
Nuxt.jsMySQLCRMSubscriptions
04 / 04
Healthcare · Scheduling
Medic Dz
Real-time medical appointment management for patients and providers, with live availability, push notifications, and mobile packaging for iOS and Android.
VueWebSocketPWACordova
Foundations
Engineering roots systems thinking.
Before building production software systems, I trained in electrical engineering, industrial informatics, automation, control systems, and applied machine learning.
2012 — 2015
B.Sc. in Electrical Engineering and Industrial Informatics
Superior National School of Technology, Algiers
Graduated with honors
Valedictorian / Major de Promotion
Highest GPA in the B.Sc. program
2015 — 2016
Additional studies in Automation and System Control
University of Science and Technology, Algiers
Completed 60/120 credits
Studied AI and machine learning applications in automation and system control
AI systems
AI in Production
AI features earn their place in production when they are reliable, observable, and designed with fallbacks. That's not a principle — it's a constraint I design from.
01
Capture real context
Experience Vault · jobs · product catalogs
↓
02
Structure the source data
Proof points · parsed jobs · normalized records
↓
03
Store reusable state
PostgreSQL · vectors · resume JSON
↓
04
Score and organize
Matching · clustering · confidence
↓
05
Generate with guardrails
LLM output · validation · deterministic fallback
What I'm strong at today
Structured context systems that preserve real experience over transient chat prompts
Human-controlled AI workflows for application answers and resume generation
Embeddings pipelines with batching, retry, and storage
Semantic clustering and vector similarity workflows
LLM-assisted product features with output validation
Deterministic fallback design for AI instability
AI-assisted engineering via Claude Code and AGENTS.md
Actively deepening
Agentic patterns and multi-step agent orchestration
RAG pipelines with retrieval quality evaluation
Production observability for LLM-powered features
“I treat AI features as production systems, not demos — with the same validation, fallbacks, and operational discipline I apply to financial workflows.”
Available for work
Let's buildsomething.
Open to consulting, technical advisory, and high-ownership engineering — especially where correctness, auditability, and AI discipline actually matter. Remote worldwide.
