›
Agentic chatbots, RAG pipelines, and full-stack apps at a US-based AI startup — Python (Django/FastAPI), React/TypeScript, LangGraph, and vector retrieval. Gold Medalist, BSCS (AI), 3.96/4.0.
I design, build, and ship production AI — agentic chatbots, RAG pipelines, and full-stack apps — end to end.
Multi-tenant LangGraph/LLM chatbots with intent classification and reliable, code-enforced conversation flows.
Vector-database retrieval (Qdrant, VoyageAI embeddings) over business knowledge with per-tenant data isolation.
Django/FastAPI microservices, REST APIs, multi-turn transactional workflows, and integrations (Jira, callbacks).
React/TypeScript frontends with Supabase, secured with Row-Level Security and end-to-end encryption.
Hybrid no-code/code pipelines (n8n + Python, Apify) and anti-bot production scrapers (Playwright, residential proxies).
Model development, fine-tuning, computer vision (YOLOv8), and deployment on Streamlit, Django, and Hugging Face.
A multi-tenant SaaS tool that automatically tests the chatbots an AI agency runs for its clients and alerts when one breaks: concurrent test runs, an LLM-as-judge issuing pass/fail verdicts with reasons, and hourly scheduled regression alert emails.
Role-based academic system where faculty submit per-course document folders against a two-phase checklist; an admin reviews submissions, flags missing items in real time, and issues auto-generated PDF review certificates.
Real-time emotion classifier — end-to-end pipeline fine-tuning ResNet50V2, VGG16, and a custom CNN, deployed as a full-stack Django app with image upload and live camera.
A RAG system for conversational Q&A over custom data: ingests uploaded files and live-scraped websites, then answers with sources via a LangChain + Gemini + FAISS pipeline.
Voice + text assistant supporting Urdu and English — real-time transcription, generative responses, and text-to-speech in a conversational UI.
End-to-end data-science project predicting customer credit risk: EDA, statistical feature engineering, and model benchmarking (Random Forest, XGBoost), deployed with a standalone executable.
A no-code AutoML web app: upload data, run EDA, and train regression/classification models through a leakage-safe scikit-learn pipeline, then download the trained model.
Detection + OCR system that locates and reads vehicle license plates from images, deployed as an interactive web app.
More projects
Production systems built at ReadyChatAI. This is proprietary company work, so there are no public repositories or live demos.
Re-architected the core microservice powering automated social-media DM conversations — LangGraph with GPT-based intent classification and deterministic, code-enforced routing.
Retrieval layer with VoyageAI embeddings and Qdrant, using per-business filtering for tenant data isolation over business knowledge.
KPI cards + Chart.js views for bot health; re-engineered metrics to a suppression-aware model, correcting reply rate from ~44% to ~56–70% and cutting false 'bot down' alerts to near zero.
Few-line embed + self-serve dashboard API, hardened with rate limiting, domain validation, and message-size limits.
Hybrid no-code + Python pipeline — Apify scraping, LLM distillation, agentic article/image generation, scheduled publishing via Metricool and auto-publish to the company blog.
Scraper covering all MercadoLibre sale and rent listings, defeating strong bot detection with residential proxies, throttling, and user-agent rotation.
React/TypeScript client with end-to-end encryption (ECDH P-256, AES-256-GCM) and Supabase RLS hardening; coordinated mobile/backend/frontend and ran QA.
ReadyChatAI LLC
Remote — Chicago, USA
Mar 2026 – Present · Junior: Nov 2025 – Mar 2026
University Institute of Information Technology (UIIT)
Feb 2023 – May 2023
↗ Certificate
AI and Backend Engineer with production experience building agentic chatbot systems, RAG pipelines, and full-stack web apps at a US-based AI startup.
I work across the stack — Python (Django/FastAPI) backends, React/TypeScript frontends, LangChain/LangGraph agentic workflows, and vector-database retrieval — and I'm comfortable with automation, web scraping, and application security (RLS, end-to-end encryption) through deployment and cross-team delivery. My foundation is in applied ML, computer vision, and NLP.
PMAS-Arid Agriculture University Rawalpindi (UIIT Campus)
Let's build
something real. →