Shipping AI at COROS · open to select freelance

Maaz Rehman

Founding Full-Stack AI Engineer at COROS AI.

0+
concurrent users in prod
0%
AI response time cut
0%
double-bookings eliminated
Karachi, Pakistan · Remotescroll
RAG
Next.js
Python
TypeScript
Pinecone
OpenAI
Llama
GCP
Docker
PostgreSQL
Redis
WebSockets
React
Node
DeepSeek
Anthropic
Microservices
Tailwind
Framer Motion
RAG
Next.js
Python
TypeScript
Pinecone
OpenAI
Llama
GCP
Docker
PostgreSQL
Redis
WebSockets
React
Node
DeepSeek
Anthropic
Microservices
Tailwind
Framer Motion
01/about

AI systems that actually ship.

I build AI systems that ship to real users — not demos. At COROS I'm the founding engineer on a 24/7 LLM-powered coaching platform: multi-provider failover, a production ElevenLabs voice clone, AI guardrails, and the iOS + Android app end-to-end. Before that I scaled PRIV's on-demand marketplace across 35+ U.S. cities. I care about latency budgets, hallucination guards, and UX that makes AI feel like a trusted advisor.

  • latency budgets
  • hallucination guards
  • zero-demo policy
  • ship weekly
02/experience

Where I’ve shipped.

presentApr 2025 — PresentRemote

Founding Full-Stack AI Engineer

0+
Concurrent Users
0%
Latency Cut
0
Platforms Shipped
  • Built COROS's 24/7 AI coaching platform from zero — a microservice backbone serving LLM-powered coaching to professionals, founders, and leaders.
  • Cut AI response latency 78% (27–32s → 6–10s) via semantic search optimization.
  • Engineered a semantic memory system with topic isolation and time-based decay, plus zero-downtime LLM failover across OpenAI, DeepSeek, Together AI, and Llama.
  • Shipped a production ElevenLabs voice clone and AI guardrails that block harmful coaching advice before delivery.
  • Shipped the iOS + Android app end-to-end — full CI/CD and store release lifecycle — while cutting GCP spend with request-based scaling.
LLM OrchestrationRAGNext.jsNodeGCPPineconeElevenLabsReact NativeCI/CD
Apr 2023 — May 2025Remote

Full-Stack Engineer

+0%
Bookings
0%
Double-Books
+0%
Mobile Conv.
  • Scaled PRIV's on-demand beauty marketplace across 35+ U.S. cities — successful bookings up 30% through reliability and UX fixes.
  • Eliminated 95% of double-bookings with real-time availability checks and optimistic locking.
  • Built a WebSocket notification system for live appointment updates — missed appointments down 40%.
  • Rebuilt the mobile-first React booking UI — mobile bookings up 50%.
ReactNodeWebSocketsPostgreSQLRedisiOSAndroid
05/stack

Tools I reach for.

maaz@stack ~ % ai/ml09 loaded

$

  • RAG & Multi-RAG
  • Multi-Agent Systems
  • RAG Eval Pipelines
  • Memory System Evals
  • Prompt & Context Engineering
  • LLM Failover & Routing

# daily drivers — llm systems, end to end — hover a chip to inspect

~/frontend02
  • React
  • Next.js
  • TypeScript
  • Tailwind
  • Framer Motion

// interfaces that feel alive

~/backend03
  • Node.js
  • Python
  • WebSockets
  • PostgreSQL
  • MongoDB
  • Redis

// where the real work happens

~/cloud-infra04
  • GCP
  • Azure
  • Docker
  • CI/CD
  • Microservices

// ship it, scale it, sleep at night

25 tools in rotation — always compiling new ones

07/education
NED University of Engineering and Technology logo
NED University of Engineering and TechnologyB.S. Computer Science — Major in Artificial Intelligence & Machine Learning2023 — 2027 · Karachi, Pakistan
08/contact

Have Something Worth Building?

Hiring, founding teams, or shipping a real AI product — happy to talk. I reply within a day.