Muhammad Safwan Ahmad Saffi

AI Engineer | Agentic Systems Specialist | 4x Hackathon Winner

Based in: Lahore, PakistanMost recently: Engineering Intern, AtlazeExperience: 1 year

AI Engineer and serial hackathon winner with a track record of building production-grade agentic systems. From developing RAG agents that cut unsupported responses by 85% at NETSOL to engineering complex SEO swarms and lead qualification pipelines, I specialize in bridging the gap between LLM research and scalable business automation. Recognized by MIT and Harvard for technical excellence, I build tools that don't just work—they scale.

Experience

Engineering Intern

AtlazeRemote, Wyoming, USAinternshipJul 2026 – Sep 2026

Built and deployed agentic AI automation systems integrating CRM and ERP platforms with LLM agents, reducing manual data entry and overhead tasks by 40% while accelerating business processing and reporting by 30%. Managed and customized core Shopify store logic to support product development and technical operations, increasing operational efficiency by 20%.

CRMERPLLM agentsShopify

AI/ML Trainee to AI/ML Intern

NETSOL Technologies PakistanLahore, PakistaninternshipFeb 2026 – Jun 2026

Designed and deployed a production LangGraph StateGraph RAG agent, scraped and normalized 378 enterprise pages, then chunking, embedding, and indexing data using LangChain, FAISS, and HuggingFace. Built grounded retrieval workflows using ChatGroq Llama 3.3 70B, advanced chunking, retrieval, and fallback strategies, reducing unsupported responses by 85% through context-grounded answers.

LangGraphStateGraphRAGLangChainFAISSHuggingFaceChatGroqLlama 3.3 70B

Data Science Track Lead

BuildablesRemoteworkAug 2025 – Nov 2025

Guided 20+ fellows through a structured Python and Data Science curriculum; improved cohort project delivery rate through structured milestone reviews.

PythonData Science

Skills

Technical

Python

Git

Agents

GitHub

RAG

LangChain

LangGraph

Tool Calling

FAISS

OpenAI

Gemini

Groq

Embeddings

Feature Engineering

Model Evaluation

Semantic Search

FastAPI

REST APIs

Web Scraping

Streamlit

CRM/ERP Automation

React

Next.js

JavaScript

Tailwind CSS

Postman

Hugging Face Spaces

scikit-learn

NLP

NumPy

Pandas

C++

MCP

Gradio

Vercel

Node.js

MongoDB

SQL

ChromaDB

React Native

TypeScript

Odoo

PyTorch

Docker

AWS

Projects

AI Lead Qualification System

Lead Engineerpersonal

Engineered a full AI scoring pipeline: Lead Input to Validation to Duplicate Check to LLM Analysis to Hot/Warm/Cold Classification to CRM Update to Sales Alert. Implemented a 100-point scoring model with lead-source multipliers (referral 1.2x, social 0.9x), a budget-missing cap at 45 pts, per-field confidence scores, and a human review queue for borderline cases (75-79 range).

n8nLLMsGoogle SheetsGmailStreamlit

View the code

Hackathon Evaluation Engine

Developerpersonal

Built and deployed a submission ranking system combining LLM evaluation with GitHub repository signals; cited in MIT MAS.665 AI Studio, NextGen Ventures, and Vals Fellowship applications.

Claude APIGitHub APIReactNext.js

Visit the site

Agentic SEO and AEO Optimization Swarm

Developerpersonal

Developed an 8-agent parallel scoring system for SEO and AEO output; settled USDC micro-payments on Arc L1 blockchain using Circle SDK.

Next.js 16GeminiCircle SDKArc L1

View the code

Production NLP Sentiment Pipeline

Developerpersonal

Trained a soft-voting ensemble of classical ML classifiers and VADER; served real-time predictions through a FastAPI REST endpoint with structured JSON responses.

scikit-learnNLTKTF-IDFVADERFastAPI

Shell.ai Hackathon 2025, SAF Prediction

Developeracademic

Developed a multi-target ML regression pipeline predicting 10 fuel blend properties from 55 input features for sustainable aviation fuel optimization using feature engineering and cross-validation.

Pythonscikit-learnPandasNumPy

View the code

Gmail Triage Automation Agent

Developerpersonal

Implemented a Gmail drafting and triage agent using the Anthropic API and MCP protocol; enforced human-in-the-loop approval before any send action.

Anthropic APIMCPPython

Salesman Management System

Developeracademic

Developed a mobile field-force management app with real-time GPS check-in/check-out, role-based access workflows, and a live supervisor dashboard.

React NativeNode.jsGPS

Education

B.S. Computer Science

National Textile UniversityComputer ScienceApr 2022 – Aug 2026

Thesis. Salesman Management System

Coursework

OOPDatabase SystemsOperating SystemsStatisticsCompiler DesignNetworkingMachine Learning

Recognition

Awards

HEC NSCT 2026 Top Ranker

Higher Education Commission (HEC)2026

Scored 85/100, Grade A (Excellent), ranking in the top 0.4% of 33,038 candidates nationwide across AI/ML, DSA, Software Engineering, Databases, Networking, and Cyber Security.

Super Contributor Badge

Hacktoberfest 20252025

Earned by shipping code and resolving issues across 5 distinct open-source projects.

Harvard CS50x Puzzle Day Winner

Harvard University

2x Intel Hackathon Winner

Intel

2x National Hackathon Winner

National WordPress Hackathon

Includes the national WordPress Hackathon.

Certifications

Stanford Code in Place, CS106A

Stanford University2026

AWS AI Practitioner

Udacity2026

Postman API Fundamentals Student Expert

Postman

Forward Program

McKinsey.org

AI Agents Fundamentals

Unknown

NIAI Program Cohort I

NETSOL

Social impact & activities

Judge

HEC GenAI Hackathonmentor

Evaluated 330+ teams at HEC GenAI Hackathon.

Mentor

USAIImentor

Mentored at USAII (130+ countries, 808 teams).

Mentor

GenAI Genesis Canadamentor

Mentored 1,600+ hackers.

Mentor

UIUC Cozad Pitch Nightmentor

Mentored at $100K program.

Shell.ai 2025

Shell.ai 2025other2025

Competed in international hackathon.

NASA Space Apps

NASA Space Appsother

Competed in international hackathon.

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

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

More people on Kanvis

Browse more people Create your own Kanvis