AI-powered outreach system for lead extraction, enrichment, CRM workflows, and automated LLM-based reply handling.

Built an AI-powered outreach system integrating LinkedIn lead extraction, API-driven lead enrichment, GoHighLevel CRM workflows, and multi-channel outreach across LinkedIn and email. The pipeline includes LLM-based automated email reply handling for inbound leads, reducing manual operations while supporting scalable campaign execution.

Lead routing and follow-up automation for faster real estate response workflows.

Developed a real estate automation system that routes incoming leads to the right agents, triggers timely follow-up sequences, and coordinates scheduling workflows. The setup reduces response delays, standardizes CRM operations, and improves conversion outcomes.

OCR pipeline that extracts financial data and powers structured KPI tracking.

Implemented an OCR-driven data pipeline to capture structured information from receipts and financial documents, then persist it in a query-ready database layer for reporting. The system enables reliable KPI monitoring and faster analytics workflows.

Reusable integration layer enabling reliable automation across tools and platforms.

Engineered a centralized automation infrastructure that standardizes API and webhook integrations across business systems. The layer enables secure, event-driven data exchange and cross-platform workflow orchestration for scalable automation delivery.

AI generates content calendars, creates posts, and auto-publishes across 3 platforms.

An end-to-end automation system where a user provides their brand kit and business website. The AI then generates a 1-2 month content calendar, creates static content for each post, and schedules and publishes automatically across Facebook, Instagram, and LinkedIn.

Unified AI voice and chat communication system for FAQ handling and automated appointment booking.

Implemented an AI communication layer combining a GoHighLevel AI Voice Agent, an AI appointment booking chatbot, and an Auxth knowledge base for FAQ resolution. The system automates inbound conversations, captures intent, and routes qualified interactions into CRM scheduling workflows.

Beginner-friendly full-stack product manager with a FastAPI backend and React frontend.

Built a product inventory manager that demonstrates typed FastAPI routes, Pydantic request and response models, CRUD endpoints, query filtering, HTTP error handling, and a React interface connected through the Fetch API.

English and Urdu fake-news classification pipeline with preprocessing, evaluation, robustness checks, and reports.

Created a complete NLP assignment package with raw and cleaned datasets, preprocessing examples, contextual embedding outputs, adversarial analysis, trained traditional and deep models, generated figures, reports, slides, and a Colab notebook.

Interactive toxicity moderation demo that analyzes comments and simulates live chat filtering.

Built a Streamlit app that loads trained classical NLP models, cleans user text, predicts toxicity, compares confidence scores, and replays a chatroom where unsafe messages are visually removed in real time.

Subject-aware EEG classification project comparing handcrafted ML features against a patch-based Transformer.

Implemented a biomedical signal classification pipeline for five EEG brain-state classes using subject-aware splitting, frequency-domain feature extraction, Random Forest baselines, and a lightweight Transformer trained on raw time-series signals.

Text-based multi-agent system that routes university queries to the most relevant specialist agent.

Implemented a query-resolution system where user requests are vectorized, compared against agent competency descriptions using cosine similarity, and routed to the best-matching agent for an intent-specific response.

Image text extraction workflow with coordinate selection, OCR detection, and analysis utilities.

Built a computer-vision project around extracting text from images, selecting regions of interest, detecting text, and analyzing extracted content through Python scripts and a dashboard-style app.