Hamza Ahmad

AI Lead & Researcher | Building Autonomous Agents & Brain-Computer Interfaces

Based in: Lahore, PakistanMost recently: Technical Lead, Cloudpacer (Neblo AI)Experience: 3 years

Hamza is a high-caliber AI Engineer and Researcher with a rare blend of production experience (Cloudpacer, Uplift AI) and top-tier academic success (MICCAI, CHI). He has demonstrated the ability to scale autonomous agents handling 10k+ daily messages and build complex BCI frameworks like CATVis, achieving 27% better accuracy than SOTA. His technical stack is deep (PyTorch, Docker, AWS, Diffusion Models), and he shows strong product intuition in optimizing for cost and reliability.

Experience

Technical Lead

Cloudpacer (Neblo AI)Lahore, PakistanworkSep 2025

Summary

AI-Powered Freight Platform: Led development of autonomous AI agent that handles carrier negotiations, load matching, and broker communications, reducing broker response time from 2+ hours to under 30 seconds and enabling 24/7 operations without human intervention. Multi-Channel Messaging Infrastructure: Built unified platform supporting Telegram, WhatsApp, and Email, processing 10,000+ daily messages with 99.5% delivery reliability and automatic failover, enabling brokers to manage carrier relationships across all channels from a single dashboard. Technical Architecture: Owned end-to-end system design across 3 microservices (AI broker, messaging, backend), making key decisions on model selection, data flow, and scaling strategy that reduced infrastructure costs by 40% while improving throughput 3x. Cross-Functional Leadership: Collaborated with product and operations teams to translate freight industry workflows into AI agent behaviors, resulting in successful pilot deployments with 15 brokerage companies.

What I did

Developed email classification (Ignore, Updates, Opportunities) and automated draft generation systems for dispatchers

Results

Reduced broker response time from 2+ hours to under 30 seconds
Enabled 24/7 operations without human intervention
Achieved 99.5% delivery reliability for messaging infrastructure
Resulted in successful pilot deployments with 15 brokerage companies
Automated redundant dispatcher tasks by generating AI drafts indistinguishable from human-sent messages
Optimized system efficiency by containerizing a logistic regression model to pre-classify 'ignore' emails, bypassing the LLM for non-essential traffic

AI AgentsMicroservicesTelegram APIWhatsApp APIEmailModel SelectionDockerLogistic Regression

Founding AI Engineer (Y Combinator S25)

Uplift AIHybridfounderJun 2025 – Jul 2025

Summary

Voice Data Collection Platform: Architected React Native app with AWS infrastructure (API Gateway → Lambda → S3) using presigned URLs and DynamoDB metadata storage, successfully collecting 125,000+ voice notes for AI model training.

Results

Successfully collected 125,000+ voice notes for AI model training

React NativeAWSAPI GatewayLambdaS3DynamoDBPresigned URLs

Associate Software Engineer

TkxelLahore, PakistanworkJul 2024 – Jun 2025

Summary

AI Like Me Platform: Spearheaded development of personalized influencer digital twin platform using Grok LLM integrated with ElevenLabs voice synthesis and HeyGen avatars, achieving 60% influencer onboarding rate and driving 35% MRR increase. Social Media Recommendation Engine: Designed hybrid recommendation system combining content-based filtering and collaborative filtering techniques that increased user engagement by 25% and improved content discovery rates by 15% across 50,000+ active users.

Results

Achieved 60% influencer onboarding rate
Drove 35% MRR increase

Grok LLMElevenLabsHeyGenRecommendation SystemsContent-based filteringCollaborative filtering

Skills

Technical

Python

JavaScript

PyTorch

HuggingFace

OpenCV

Transformers

React Native

AWS

Git

C++

Scikit-learn

Redis

SQL

MongoDB

PostgreSQL

TensorFlow

Firebase

Docker

Keras

Projects

CATVis: Context-Aware Thought Visualization

Lead Researcheracademic

A framework for reconstructing high-fidelity images from noisy EEG brain signals by mapping brain activity into a textual feature space and utilizing diffusion models.

PythonPyTorchDiffusion ModelsEEG ProcessingNLP

Education

Bachelor of Science

Forman Christian College (A Chartered University)Computer ScienceGPA 3.9Aug 2020 – May 2024

Coursework

Data StructuresMachine LearningComputer VisionDeep LearningSoftware Engineering

Societies & activities

Teaching Assistant: Data StructuresTeaching Assistant: Machine LearningTeaching Assistant: Linear Algebra

Recognition

Awards

Swami Ram Teerath Medal

Forman Christian College (A Chartered University)2024

Awarded for highest CGPA in Faculty of Computer and Mathematical Sciences

Patents & publications

CATVis: Context-Aware Thought Visualization

publication2025

Orchestrated development of novel 5-stage EEG-to-image framework achieving 27% higher classification accuracy and 37% FID reduction over state-of-the-art methods through supervised learning and cross-modal alignment.

Mitigating Toxicity in Social Media

publication2025

Published in CHI 2025 Extended Abstracts (EA '25).

Social impact & activities

CATVis: Context-Aware Thought Visualization

MICCAI 2025conference talkJul 2025

Presentation of EEG-to-image framework research.

Mitigating Toxicity in Social Media

CHI EA '25conference talkApr 2025

Presentation of toxicity mitigation framework research.

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

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

More people on Kanvis

Browse more people Create your own Kanvis