Waqas Yousaf

AI Engineer specializing in production-grade agentic systems and RAG pipelines.

Based in: Lahore, PakistanCurrently: AI Engineer, FlickensExperience: 7 years

AI Engineer with over 2 years of experience building production-grade agentic systems and RAG pipelines, complemented by 6 years of software engineering expertise. He specializes in LangGraph, FastAPI, and reliability engineering to deploy secure, high-performance AI automation workflows.

Experience

AI Engineer

FlickensworkAug 2024 – Present

Summary

Architect and deploy production-grade AI agents, RAG engines, and workflow automations using LangChain, LangGraph, n8n, and FastAPI. Integrate custom Model Context Protocol (MCP) servers and LLM APIs (OpenAI, Claude, Groq) into client systems to automate business workflows and document processing.

What I did

Automated social media posting workflows that convert articles into videos and post them to social media platforms.
Integrated the chatbot with the website to enable interaction through FastAPI.
Managed multi-turn session states by handling sessions based on user ID to store history.
Implemented streaming responses within the FastAPI backend.

LangChainLangGraphn8nFastAPIModel Context Protocol (MCP)OpenAIClaudeGroqPython

Android Developer

Hazel MobileworkSep 2019 – Mar 2024

Summary

Engineered native Android applications using Kotlin, MVVM, Clean Architecture, and Coroutines. Integrated Firebase services, REST APIs, Google Billing, and AdMob across applications reaching over 10M+ combined downloads.

Results

Reached over 10M+ combined downloads across integrated applications.

KotlinMVVMClean ArchitectureCoroutinesFirebaseREST APIsGoogle BillingAdMob

Skills

Technical

LangGraph

LangChain

AI Agents

Prompt Engineering

FastAPI

REST APIs

Kotlin

MVVM

Clean Architecture

Python

Git

GitHub

Semantic Search

Java

Model Context Protocol (MCP)

Tool Calling

Jetpack Compose

LangSmith

Conversational Memory

Model Fallback Routing

Prompt Injection & PII Guardrails

ChromaDB

PostgreSQL (pgvector)

n8n

Webhooks

Process Automation

Context/Token Optimization

Vector Databases

Redis

Redis Caching

Rate Limiting

LLM-as-a-Judge

Response Evaluation

Distributed Tracing

Cross-Encoder Reranking

Docker

Tool Orchestration

Structured Outputs

Whisper

Pydantic

Multi-step Tool Sequencing

LiveKit

Data Validation

Projects

Real-Time Voice AI Customer Support System

Lead Developerpersonal

Building an ultra-low-latency real-time voice support agent for an electronic store using LiveKit WebRTC for bi-directional streaming, AssemblyAI for speech-to-text, Groq for intent processing, and Cartesia for natural voice synthesis. Integrating custom MCP tool-calling servers to automatically log customer complaints mid-conversation into an Excel spreadsheet and dispatch dynamic email alerts to the support team for human escalation.

LiveKitAssemblyAIGroqCartesiaModel Context Protocol (MCP)PythonExcel IntegrationWebRTC

Production RAG Agent with Model Fallback, Caching & Security Guardrails

Lead Developerpersonal

Built a LangGraph ReAct agent over a PostgreSQL/pgvector retrieval pipeline with MMR search and cross-encoder reranking for high-precision, source-attributed context. Implemented automatic model fallback (Groq ↔ DeepSeek) with a shared checkpointer so conversation state persists seamlessly across provider failover. Cut per-conversation token cost via history trimming and a Redis-backed response cache, and hardened the agent with rate limiting and input/output security guardrails (prompt-injection detection, PII masking). Instrumented the retrieval → rerank → generation pipeline with LangSmith tracing for per-stage latency, cost, and grounding visibility.

LangGraphLangChainGroqDeepSeekPostgreSQL (pgvector)RedisLangSmithMMR searchCross-encoder reranking

RAG Knowledge Assistant & Evaluator

Architectpersonal

Architected a document-based Retrieval-Augmented Generation (RAG) system utilizing vector search and dynamic state routing. Built an automated evaluation pipeline in LangSmith using an LLM-as-a-judge approach to measure faithfulness and context relevance.

LangGraphLangChainFastAPIChromaDBLangSmith

AI Chat Assistant with Semantic Memory

Backend Developerpersonal

Developed a conversational backend using pgvector to store and query message embeddings for long-term context retrieval. Optimized token usage while maintaining context continuity over extended multi-turn conversations.

PostgreSQLpgvectorLangChainFastAPI

MCP Appointment Management Agent

Engineerpersonal

Engineered an appointment scheduling agent using a custom MCP client/server architecture to execute direct calendar operations.

Model Context Protocol (MCP)n8nGroqCalendar API

Law Firm AI Front Desk Engine

Developerpersonal

Automated client intake inquiries and appointment booking with dynamic calendar checking and email notifications.

n8nAI AgentsGmail APIGoogle Calendar API

LangGraph ReAct Agent at Pattern

personal

Built a ReAct agent with model fallback between Groq and DeepSeek to handle service outages, such as OpenAI link downs.

LangGraphGroqDeepSeek

Education

Bachelor of Science in Computer Science (BSCS)

Lahore Garrison UniversityComputer Science

Hire WaqasReach out about a role, a contract or a conversation.For recruiters

Waqas's twin is AI, it can make mistakes.

More people on Kanvis

Browse more people Create your own Kanvis