Hamza Ahmad

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

Overview

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.

Lahore, Pakistan

Experience

Sep 2025

WORK

Technical Lead

Cloudpacer (Neblo AI)

Lahore, Pakistan

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.

•Led development of autonomous AI agent that handles carrier negotiations, load matching, and broker communications

•Built unified platform supporting Telegram, WhatsApp, and Email, processing 10,000+ daily messages with 99.5% delivery reliability and automatic failover

•Owned end-to-end system design across 3 microservices (AI broker, messaging, backend)

•Collaborated with product and operations teams to translate freight industry workflows into AI agent behaviors

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

Key Achievements

→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

→Reduced infrastructure costs by 40% while improving throughput 3x

→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

→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

→Reduced infrastructure costs by 40% while improving throughput 3x

→Resulted in successful pilot deployments with 15 brokerage companies

AI AgentsMicroservicesTelegram APIWhatsApp APIEmailModel SelectionDockerLogistic Regression

Jun 2025 – Jul 2025

FOUNDER

Founding AI Engineer (Y Combinator S25)

Uplift AI

Hybrid

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.

•Architected React Native app with AWS infrastructure (API Gateway → Lambda → S3) using presigned URLs and DynamoDB metadata storage

Key Achievements

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

React NativeAWSAPI GatewayLambdaS3DynamoDBPresigned URLs

Jul 2024 – Jun 2025

WORK

Associate Software Engineer

Tkxel

Lahore, Pakistan

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.

•Spearheaded development of personalized influencer digital twin platform using Grok LLM integrated with ElevenLabs voice synthesis and HeyGen avatars

•Designed hybrid recommendation system combining content-based filtering and collaborative filtering techniques

Key Achievements

→Achieved 60% influencer onboarding rate

→Drove 35% MRR increase

→Increased user engagement by 25%

→Improved content discovery rates by 15% across 50,000+ active users

Grok LLMElevenLabsHeyGenRecommendation SystemsContent-based filteringCollaborative filtering

Sep 2024 – Mar 2025

WORK

Research Assistant

Computer Vision and Graphics Lab - LUMS

Lahore, Pakistan

Brain-Computer Interface Research: Pioneered development of AI system that reconstructs images from brain signals using deep learning, achieving 27% higher accuracy than existing methods and securing publication at top-tier MICCAI 2025 conference.

•Pioneered development of AI system that reconstructs images from brain signals using deep learning

Key Achievements

→Achieved 27% higher accuracy than existing methods by mapping EEG signals to textual feature space before image generation

→Developed a multi-stage pipeline: EEG pre-processing -> Textual Feature Mapping -> Diffusion Model Generation -> Caption Re-ranking

→Secured publication at top-tier MICCAI 2025 conference

→Achieved 27% higher accuracy than existing methods

→Secured publication at top-tier MICCAI 2025 conference

Deep LearningBrain-Computer InterfaceComputer Vision

Oct 2023 – Oct 2024

WORK

Software Engineer

Adinteractive

Remote

E-commerce Computer Vision System: Implemented YOLOv8-based product detection system for trailer advertisements with 10,000+ annotated products, achieving 40% increase in user engagement and 50% reduction in inference time through MongoDB to AWS DynamoDB migration with Redis caching.

•Implemented YOLOv8-based product detection system for trailer advertisements with 10,000+ annotated products

•Migrated from MongoDB to AWS DynamoDB with Redis caching

Key Achievements

→Achieved 40% increase in user engagement

→Achieved 50% reduction in inference time

YOLOv8Computer VisionMongoDBAWS DynamoDBRedis

Project Portfolio

ACADEMIC

Present

CATVis: Context-Aware Thought Visualization

Lead Researcher

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

Aug 2020 – May 2024

Bachelor of Science

Forman Christian College (A Chartered University)Specialization in Computer Science

GPA: 3.9

Relevant Coursework

Data StructuresMachine LearningComputer VisionDeep LearningSoftware Engineering

Societies & Activities

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

Impact & Recognition

Awards

Forman Christian College (A Chartered University) · 2024

Swami Ram Teerath Medal

Awarded for highest CGPA in Faculty of Computer and Mathematical Sciences

Patents & Publications

2025 · PUBLICATION

CATVis: Context-Aware Thought Visualization

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.

2025 · PUBLICATION