I build AI agents
and the platforms they run on.
AI and full-stack engineer. I build agentic systems that reason, call tools and take real action, plus the RAG pipelines and production APIs behind them. Six years shipping products in React, Next.js and Node, at scales up to 100M+ users.
agents: this site is an MCP server: haroonjawad.com/api/mcp
What I can build for you
Five things I take on. Each one is scoped, with concrete deliverables, not a retainer for vague “AI strategy”.
Agentic AI systems
Project-based or fractional lead engineerDesign and build LLM agents that reason, plan, call tools, and take action in production, not demos. LangGraph/LangChain orchestration, tool-calling, evaluation, and cost control.
- Agent architecture and orchestration graph
- Tool/function-calling layer against your real systems
- Observability, tracing, and evaluation harness
- Deployment with cost and latency budgets
RAG pipelines over your own data
Project-basedRetrieval-augmented systems that give grounded, cited answers over unstructured document sets. Chunking strategy, embeddings, pgvector, retrieval quality measurement.
- Ingestion and chunking pipeline
- Vector store setup (pgvector/Supabase) and retrieval tuning
- Grounded answer layer with citations
- Retrieval-quality benchmarks before/after
AI-native full-stack product build
Project-based or ongoingEnd-to-end product delivery where the AI is the product: Next.js/TypeScript front end, Node/NestJS services, PostgreSQL, auth, billing, and the model layer wired together.
- Production Next.js + TypeScript application
- API/services layer and data model
- Auth (OIDC/OAuth 2.0/WebAuthn) and billing integration
- CI/CD and infrastructure
AI cost audit
Fixed scope, roughly two weeksA fixed-scope review of what your LLM features actually cost to run. Most teams are paying several times more than they need to, and the causes are boringly consistent: whole conversation histories replayed on every call, prompt caching that looks enabled but never hits, and the most expensive model doing work a cheap one does just as well. Applying exactly this took one production bill down by ~80%.
- Where the spend actually goes, broken down by feature and call path
- Prompt-caching audit, what invalidates the cache and the fix
- Model-routing plan: which tasks belong on which tier
- Prioritised changes, each with its projected saving
- Before/after monitoring so the saving is verifiable, not claimed
Agent-readiness & AEO
Fixed scope, or ongoing alongside content workTwo sides of one problem: being usable by AI agents, and being recommended when someone asks ChatGPT, Claude or Perplexity instead of Google. Neither is SEO. Answer engines quote sources they can parse, verify and attribute, and agents want a tool call rather than a page to scrape, so the work is structuring your capability and content for machine readers. This site is the worked example.
- Audit of how agents and answer engines currently see, quote and attribute you
- MCP server exposing your product as typed tools, plus registry listing
- llms.txt, structured data and machine-readable JSON endpoints
- Content restructured into extractable, citable answers
- Re-check afterwards, so the change is measured rather than assumed
Available for AI and full-stack engineering work, agentic systems, RAG pipelines, and production product builds. Takes on a focused number of engagements at a time, so each client gets senior attention rather than a slice of it. Engagements typically start with a short scoping call and end with something shipped and handed over. Also open to senior and lead AI engineering roles.
Things I've shipped
Products I architected and built end to end, not tutorials, not side experiments that never left localhost.
This site, as an MCP server →
Live · 4 MCP toolsRebuilt haroonjawad.com so an AI agent can use it directly rather than scraping it. Four MCP tools over streamable HTTP (résumé, services, availability, and a brief-intake tool that reaches me), plus a machine-readable JSON résumé and an llms.txt for discovery. One source of truth feeds the tools, the endpoints and the pages, so nothing drifts. The assistant on the site calls those same tools and shows each call as it happens.
wctx
Open source · local-first MCP serverLocal MCP server that gives a coding agent in one repo access to what a session in a related repo already learned, root causes, decisions, dead ends, with per-repository commit provenance and a staleness check so history is never mistaken for current truth. Repo stays the edit boundary; workspace becomes the context boundary.
Multi-Agent Orchestration →
Open source · 5 models, 6 phasesFive AI models running as a team with separated duties, building a Notion clone unattended across six phases: an orchestrator, a backend dev, a frontend dev, QA, and an adversary whose only job was to break the app. Roles were described in prompts but enforced with file permissions, developers couldn't edit the defect ledger, QA couldn't touch product code, and no model could certify its own fix. Produced 26,132 lines of TypeScript and a 14-entry defect ledger, 13 of which the adversary found.
Analytra ↗
Live SaaS · built and shipped soloAI-powered analytics SaaS with natural-language querying over 11 live data sources (ad platforms, CRMs, Stripe, web analytics), plus automated AI-generated reports with PDF export. Built solo.
AI Interchange
Flywheel AI (private)Shared multi-tenant AI platform powering multiple AI products: centralized billing (Stripe, custom credit wallets), cross-application authentication, a multi-model AI chat app, and a master user system letting one account move across apps under a single sign-on and shared wallet.
Workout Form Coach
Personal project · 6 months in the makingReal-time exercise form correction app. Trained an LSTM model on MediaPipe pose-tracking data to detect exercises (bicep curls, squats, pushups) and score form against joint-angle ranges defined with fitness experts, with live feedback and workout summaries.
This portfolio is an MCP server.
Point Claude, ChatGPT or any MCP client at haroonjawad.com/api/mcp and it can read my résumé, check my availability, list what I take on, and send me a brief, no human in the loop. The same tools power the assistant on /chat, so you can watch the tool calls happen.
Roles I'm open to
Full-time or contract. If your opening looks like one of these, we should talk.
Full Stack Engineer
Ships the whole thing: React and Next.js front ends, Node and NestJS services, PostgreSQL and MongoDB, auth and billing wired in. Six years and 40+ delivered projects of doing exactly this.
AI Automation Specialist
Turns manual, judgement-heavy workflows into agent-driven ones, tool-calling agents, RAG over internal documents, and the evaluation and cost controls that keep them trustworthy in production.
Product Engineer
Owns a product surface end to end: scoping with stakeholders, building it, shipping it, then changing it based on what real usage says. Comfortable being the person accountable for the outcome, not just the ticket.
Forward Deployed Engineer
Sits with the customer and makes the platform work in their environment. Has done this for Google, CVS Health and Uber, integrations, custom workflows, and turning what breaks in the field into product changes.
An engineer who works with AI daily
I build agentic AI systems, the kind that actually do something. Tool-calling agents, RAG pipelines that give cited answers over your own documents, and the production APIs and front ends that make them usable. I care about the unglamorous half: evaluation, observability, cost control, and what happens when the model gets it wrong.
I've delivered 40+ projects for clients across e-commerce, healthcare, marketing and edtech, reaching Top Rated Plus on Upwork with a 100% job success score. Lately that's meant building AI products end to end on my own, like Analytra, which answers questions in plain English across 11 live data sources.
Day to day I lead engineering on a privacy-preserving identity platform serving 100M+ users, that's where the depth in authentication, browser cryptography and regulated production systems comes from. I also write about this work: recent essays on agent deployment security and hiring for AI-collaborative engineering reached 13,500+ and 8,800+ readers.
Technologies I work with daily
Front end, back end, and the model layer. What I actually build with.
Where I've built things
The short version. Full history, including every highlight per role, is in the résumé JSON.
Lead Full Stack Engineer · Private Identity
May 2025 – PresentEnterprise privacy-preserving identity verification; biometric auth via WebAssembly. Lead and mentor a web engineering team of 8, owning architecture decisions and cross-functional delivery across PrivateID's identity verification platform.
- Architected PrivateID's Orchestration Engine, enabling configurable workflows, dynamic rule evaluation, and automated multi-step verification flows that reduce manual intervention and accelerate customer integrations.
- Built reusable SDKs for face login, liveness detection, enrollment, and deletion; shipped ML-driven features (age estimation, spoof detection, risk scoring) and LLM-powered explanation systems that interpret verification outcomes and guide users through successful retries.
Senior Full Stack Engineer · Private Identity
Sep 2022 – May 2025Built and maintained the Ultrapass Web SDK, WebAssembly-based biometric authentication with Fully Homomorphic Encryption (FHE). No biometric data ever leaves the device; GDPR and CCPA compliant.
- Shipped the Ultrapass Web SDK backed by a NestJS microservices orchestration layer and a Node/Express/MongoDB backend with native C++ module bindings.
- Implemented OIDC/OAuth 2.0 (PKCE + Authorization Code Grant) for biometric auth flows.
React Developer · Atentiv LLC
Sep 2021 – Sep 2022ADHD digital therapeutic platform. Built and maintained the Patient Management Portal with full role-based access control.
- Built the Patient Management Portal with React, Redux, Next.js, NestJS, TypeORM, and PostgreSQL, with full RBAC across patient, caregiver, physician, and admin roles.
- Implemented real-time D3.js data visualization for patient monitoring, delivered as reusable React components powering the platform's analytics.
JavaScript Developer · Jslytics
Mar 2020 – Sep 2021Analytics and marketing technology. Built server-side tagging and data-transformation solutions for enterprise clients.
- Built server-side Google Tag Manager tagging solutions and custom JavaScript variables to capture and transform user data before it reached analytics platforms.
- Configured consent management for GTM deployments, keeping client-side data collection compliant across regions.
Full Stack Developer · Freelance (Upwork)
Sep 2019 – Mar 2021Top Rated Plus · 100% Job Success Score · $100K+ earned · 44 jobs · 3,396 hours.
- Delivered 40+ projects, including several large-scale enterprise engagements (e.g., Bondbury), across e-commerce, healthcare, marketing, and edtech using React, Next.js, Node.js, TypeScript, Firebase, and MongoDB.
- Achieved Top Rated Plus status (top 3% globally) with a 100% job success score.
What people say after we ship
Haroon is a great UI/UX developer. He has worked well with our distributed teams to build out the world's best ADHD digital therapeutic solution. Without his knowledge and creativity we would never have been as successful.
Haroon does excellent work and looks for alternative solutions. He is a great team player and has a positive attitude. I highly recommend him.
Great freelancer, thanks so much! Great communication and delivery.
Let's build something intelligent.
Available for AI and full-stack engineering work, agentic systems, RAG pipelines, and production product builds. Takes on a focused number of engagements at a time, so each client gets senior attention rather than a slice of it. Engagements typically start with a short scoping call and end with something shipped and handed over. Also open to senior and lead AI engineering roles. usually within 1–2 business days.