Skip to content
AF

Hi, my name is

Ahmad Faraz.

I build things for the web and AI.

I'm a Computer Science graduate from LUMS, specializing in Machine Learning and Full-Stack Development. Most recently, I worked as a Full-Stack Engineer at Tajir (YC W20), and I'm currently open to new opportunities.

View my workGet in touch

01. About Me

I'm a Computer Science graduate from LUMS, where I graduated in May 2025. My journey in tech has led me from web development to the exciting realms of AI and Machine Learning, and most recently to full-stack engineering at scale as a Full-Stack Engineer at Tajir (YC W20). I thrive on challenges and continuously seek to expand my skillset.

Ahmad Faraz

Languages

  • Python
  • TypeScript
  • JavaScript
  • Kotlin
  • SQL
  • C++
  • C
  • Haskell

Frameworks & Technologies

  • Flask
  • FastAPI
  • Node.js
  • Express.js
  • React.js
  • Redux
  • Android SDK (Kotlin, MVVM)
  • Office.js

AI & Machine Learning

  • OpenAI API
  • Anthropic Claude
  • LangChain
  • RAG
  • Vector Search
  • AI Agents
  • MCP
  • PyTorch
  • TensorFlow
  • Scikit-learn

Databases & Data

  • PostgreSQL
  • MongoDB
  • Supabase
  • BigQuery
  • Firestore
  • Pandas
  • NumPy

Cloud & DevOps

  • AWS (Lambda, S3, RDS, Cognito)
  • GCP (Cloud Run, GCS)
  • Docker
  • Terraform
  • CI/CD
  • Vercel

Developer Tools

  • Git
  • n8n
  • Retool
  • Figma
  • Jira
  • VS Code
  • Cursor
  • Claude Code

02. Where I've Worked

Full-Stack Engineer @ Tajir (YC W20)

June 2025 - July 2026 · Lahore, Pakistan

  • Led delivery-operations systems for a B2B commerce platform moving 4,500+ orders/day across 62 zones — from a handcart-batching algorithm to a rotational-driver assignment model — across Flask, PostgreSQL, Retool, and a Kotlin Android app.
  • Took the order-editing and cancellation-reason features from backend to Android UI, driving a 10x jump in structured customer feedback within 10 days of launch.
  • Tracked down a driver-pairing bug that was corrupting dispatch, cutting incidents from several a week to zero, then resolved a live production outage affecting 143 orders in under 10 minutes.
  • Found a delivery-planning failure that had silently orphaned 190+ deliveries, shipped a fix, and cut a recurring $270/month cloud cost by redesigning the analytics pipeline behind it.

03. Some Things I've Built

ReplyGenie preview

ReplyGenie

An AI-powered Outlook add-in that reads the active email thread and drafts contextual replies with GPT-4o-mini — no more tab-switching to catch up on a conversation. Built as a multi-tenant SaaS with FastAPI and Supabase Auth, with per-tenant data isolation and full draft-history logging.

  • React
  • Office.js
  • FastAPI
  • Supabase
  • OpenAI
Zero-Shot Vision and Language Navigation preview

Zero-Shot Vision and Language Navigation

A zero-shot indoor navigation agent that follows plain-language instructions with no finetuning, built for my undergraduate thesis. Combines panoramic scene summarization with a graph-based spatial memory, reaching a 47.13% success rate on Room-to-Room and outperforming prior training-based approaches.

Read the thesis report(PDF)
  • LLaVA-NeXT
  • GPT-4o
  • Matterport3D
IntelliLearn preview

IntelliLearn

An AI-powered ed-tech platform for course exploration, enrollment and personalized learning, with a GPT-3.5 chatbot, text summarization, a timed quiz system with concept feedback, and student/admin dashboards.

  • MERN
  • GPT-3.5
  • React Redux
Song Recommendation Platform preview

Song Recommendation Platform

A serverless recommendation backend on AWS Lambda and API Gateway, provisioned end-to-end with Terraform — Cognito for auth, RDS Postgres for persistence, and a React frontend on S3 + CloudFront.

  • AWS
  • Terraform
  • Python
  • React

AttendiGo

A full-stack attendance platform for teachers — class creation, a reporting dashboard, and AI-generated insights. A self-hosted n8n workflow joins attendance and student records on a webhook trigger and calls an LLM to surface trends and flag at-risk students.

  • Flask
  • React
  • Supabase
  • n8n

AI Chef Assistant

A cooking assistant that generates personalized recipes with step-by-step instructions and matching images using Llama 3 and DALL·E 2, plus nutrition insights and prep time.

  • Flask
  • React.js
  • Llama 3
  • DALL·E 2

Content Moderation & Toxicity Classification

Naive Bayes, RNN and BERT compared on toxic-comment classification, with BERT reaching 88.29% recall after class-imbalance handling.

  • Python
  • Hugging Face

US Crime Data Analysis and Prediction

Four decades of US crime and demographic data analysed with causal inference, then forecast with a two-layer neural network and SARIMAX modelling. Written up on Medium.

  • Python
  • TensorFlow/Keras
  • Scikit-Learn
  • Statsmodels

04. Get In Touch

I'm currently looking for new opportunities. Whether you have a question or just want to say hi, I'll do my best to get back to you.

Prefer email? ahmedd.fz11@gmail.com