Ahmed Hassan

AI Product Manager & ML Engineer specializing in LLM ecosystems and cross-lingual retrieval.

Based in: IslamabadCurrently: AI Product Manager, ImagineArt (Vyro AI)Experience: 1 year

Hire me

Ahmed is a Computer Science student at LUMS with a strong background in machine learning research and AI product management. He has successfully shipped global plugin ecosystems and published research on multilingual embedding alignment and medical QA benchmarks.

Experience

AI Product Manager

ImagineArt (Vyro AI)IslamabadworkJun 2026 – Present

Summary

Own, and shipped the entire ImagineArt plugin ecosystem across Canva, Shopify, Figma, and Adobe Creative Suite, taking each integration from market research and spec through build and marketplace launch. Own the ImagineArt MCP platform for AI agents and led it to publication on OpenAI, overseeing tool design, docstrings, testing, and interactive frontend widgets. Adapted MCP tools and schemas for multiple agentic harnesses, ensuring reliable tool-calling across different agent frameworks. Built Atlas, a fully autonomous internal social media agent that turns prompts into static and video posts, writes captions, and schedules posting across all major platforms using a Hermes Agent and LLM scripting.

What I did

  • Tracked plugin usage and GTM events via Mixpanel for ChatGPT/MCP integrations and internal databases for Canva and Shopify.
  • Tracked the ImagineArt ecosystem through Mixpanel and internal database events to monitor integration performance.
CanvaShopifyFigmaAdobe Creative SuiteMCPOpenAIHermes AgentLLM scriptingMixpanel

Machine Learning Engineer

InstantAI LabsRemoteworkJul 2025 – May 2026

Summary

Researched and built a low-resource dataset for Cross-Lingual RAG; aligned multilingual sentence embeddings via Procrustes transformations, boosting retrieval accuracy by 30% across multiple LLMs.

What I did

  • Fine-tuned models to fit specific use cases.
Cross-Lingual RAGProcrustes transformationsLLMsFine-tuning

Engineering Intern

Rayn GroupIslamabadinternshipJun 2025 – Aug 2025

Summary

Engineered an XGBoost-based predictive dashboard for malnutrition hotspot detection (>85% accuracy), integrating real-time geolocation data pipelines into a production full-stack web application with interactive map visualizations.

Results

  • Achieved >85% accuracy for malnutrition hotspot detection.
  • Received Senior Engineer Recommendation.
XGBoostGeolocation data pipelinesFull-stack web applicationMap visualizations

Skills

Technical

PythonPython
PyTorchPyTorch
LangChain
TypeScriptTypeScript
JavaScriptJavaScript
Next.jsNext.js
ReactReact
GitGit
LinuxLinux
LaTeX
TensorFlowTensorFlow
FastAPIFastAPI
FlaskFlask
Node.jsNode.js
MongoDBMongoDB
PostgreSQLPostgreSQL
XGBoost
LangGraph
scikit-learn
Databricks
AWS
GCPGCP
C/C++C/C++
PySpark
DockerDocker
Azure
Haskell
Fine-tuning
Mixpanel

Languages

English
Urdu

Projects

HUDI - Housing & Urban Development Index

Lead Developeracademic
Built a three-dimensional fine-resolution urban liveability index for Lahore across 29,492 grid cells (250 m) and 21 months (Jan 2021 - Feb 2026), fusing 15 open satellite and OSM data layers
GeoPandasSentinel-1/2TROPOMIOSM data

Cross-lingual retrieval and low-resource datasets research

personal
Published research at EACL and LREC-COLING regarding cross-lingual retrieval and low-resource datasets.
Procrustes transformation

SIGUL Dataset Alignment

Researcher / Engineerpersonal
Constructed the pipeline, finalized dataset selection, and ran experiments to obtain final results for the SIGUL paper.

FedMosaic - Federated LLM Adapter Routing

Lead Engineerpersonal
Engineered a federated learning framework for heterogeneous LLMs using Block-wise PQ-LoRA and Layerwise Routing to optimize cross-client communication efficiency. Developed Data-Free Geometric Alignment to align client embedding subspaces without sharing raw data.
PyTorchPEFTBitsAndBytesBlock-wise PQ-LoRALayerwise Routing

Music Chatbot

personal
Fine-tuned a RAG-based LLM for a music chatbot as part of client work at InstantAI Labs.
RAGLLM

Education

Bachelor of Science

Lahore University of Management Sciences (LUMS)Computer ScienceAug 2022 – May 2026

Coursework

Advanced Topics in MLMLOpsDeep LearningSoftware EngineeringDatabases

Societies & activities

Director Marketing, LUMS Community Service Society

Recognition

Awards

Dean's Honor List

Lahore University of Management Sciences (LUMS)2024

Patents & publications

Beyond Fine-Tuning: Procrustes Alignment of Multilingual Embeddings for Low-Resource Cross-Lingual Retrieval

publication2026

SIGUL 2026 @ LREC-COLING 2026. Engineered a computationally efficient Procrustes transformation to align Urdu-English embedding subspaces in pre-trained models.

U-MIRAGE: Benchmarking Chain-of-Thought Reasoning for Urdu Medical QA

publication2026

AbjadNLP @ EACL 2026 (Oral Presentation). Built a bilingual Medical QA benchmark (7,663 items) evaluating CoT reasoning and live-RAG pipelines across 6 LLMs.

Social impact & activities

Director Marketing

LUMS Community Service Societyorganizer

Oral Presentation: U-MIRAGE

AbjadNLP @ EACL 2026conference talkFeb 2026

Presented research on Urdu Medical QA benchmarks.

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

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