Omer Haydar

Omer Haydar

AI Research Engineer | Technical Sales & Solutions Engineering Aspirant

Based in: Lahore, PakistanCurrently: AI Research Engineer, VEON Ltd.Experience: 4 years

Hire me

Technical leader and researcher bridging biology and AI, now pivoting into Technical Sales and Solutions Engineering. Expert in translating complex ML architectures (protein folding, emotion detection) into business value. Proven track record in building AI departments, securing corporate sponsorships, and driving user engagement through data-driven strategies.

Experience

AI Research Engineer

VEON Ltd.UAEworkAug 2026 – Present
  • Developed front-ends for specific modules including a multi-tenant data-specific chat feature with tools for insight generation, report building, and survey building.
React

Lead Analytics and AI Engineer

SHAPE Global Ltd.workJan 2025 – Present

Summary

Established the AI/ML department from the ground up, focusing on team building and high-energy recruitment.

What I did

  • Scoped and shipped new offerings (longitudinal, case, and causal-inference studies) with predictive features.
  • Partnered with product, design, and research to translate non-technical expectations into BRDs, functional requirements, and user stories.
  • Set up knowledge base for RAGs, knowledge graphs, and agentic workflows.
  • Built a population data-simulation engine using Big-5, World Values, and LLM injection to stress-test survey coverage.
  • Managed Azure access & deployment governance including RBAC, quotas, CI/CD, AKS, and Kubernetes.
  • Developed a multi-tier PAYG pricing model with token-based wallets adopted as the GTM framework.
  • Built an in-house multi-tenant collaborative project tracking system and internal tools such as a case studies engine.
  • Developed user-facing workflows from design to development.
  • Used TypeScript to manage complex multi-tenant structures, access roles, and parameter restrictions for internal tools.
  • Used fine-tuned BERT models and small LLMs to standardize best practices and survey questionnaire generation.

Results

  • Enhanced approval ratings by 31% through causal-inference and longitudinal studies.
  • Achieved SOTA (80-84%) synthetic-survey fidelity underpinning index defendability.
  • Cut index development time from months to weeks via automated data lifecycles.
  • Reduced index development time from months to weeks via automated data lifecycles.
  • Established the AI/ML department and led recruitment.
RAGAgentic workflowsCausal inferenceAzure AI FoundryArgoCD.NETSQLGTM frameworkAzure AKSKubernetesPower BIReactTypeScriptKerasPyTorchTensorFlowUnsloth

Co-founder

GiftShiftwork2025 – Jan 2025

Summary

Co-founded an AI-enabled gifting platform focused on personalized recommendations.

What I did

  • Launched an AI-enabled B2C gifting platform; developed business strategy, fulfillment pipelines, and an ML-based recommendation engine for artisanal vendors.
  • Built a custom Shopify frontend with custom backend injection.
PyTorchTensorFlowHugging FaceWeights & Biases

Skills

Technical

Knowledge graphs
Data Science
Protein Folding
LLMs
Hugging Face
Deep Learning
BERT
Azure AKS
KubernetesKubernetes
PyTorchPyTorch
Intro to AI & ML
TensorFlowTensorFlow
Weights & Biases
Power BI
RAG
Data Structures
Object-Oriented Programming
ICD-9/10
CPT code analysis
Wav2Vec2
ASR
VLM
UNET
Diffusion-based edge enhancement
fMRI analysis
Linear Algebra
Computational Biology
GDPR compliance
Genetics
Molecular Biology
MLOps
Microsoft AI Foundry
AWS Bedrock
Node.jsNode.js
FastAPIFastAPI
Azure DevOps
GCPGCP
VLMs
Agentic workflows
Causal inference
Ablation studies
Model benchmarking
PythonPython
ReactReact
TypeScriptTypeScript
Fine-tuning
Ensemble methods
Synthetic data generation
Vector search
Unsloth

Areas of expertise

ROI quantification

Projects

Genomic Risk Prediction

External Associateacademic2025 – Present
EDA on 10K+ pathogen genomes vs. AMR phenotypes.
Data ScienceNumpy

EmoMosaic

Research AssociateacademicNov 2024 – Present
Fine-tuned BERT on cross-domain emotions in text datasets to create a generalized emotion detection model. Achieved 53% performance (SOTA) compared to single-dataset models. Model card: vtlustos/EmoMosaic-large Finetuned.
PythonLLMsBERTFine-tuningQuantizationDeployment
View the code

Low-Compute Protein Structuring Model

personal
Developed an optimized protein structure model integrated with AlphaFold for 3D structuring, focusing on compute efficiency. It used negative target assembly with established drug carriers to create the structure of protein. It used a simple attention for whole sequences and using Bayesian predictability for synthetic data.
PyTorchBayesian InferenceTransformersPythonDeep Learning

Emotion Recognition for InterSpeech 2025

Research Associateacademic
Fine-tuned an advanced ASR model to detect and classify human emotion from natural audio, contributing to a submission for a leading international AI conference.
PythonAWSSagemakerLocal TrainingDistributed training and inferenceEvaluation
View the code

Education

B.S.

Lahore University Of Management SciencesBiology (CS Coursework)2025

Coursework

Data ScienceDeep LearningAI/MLOOPData StructuresLinear AlgebraProbability

Recognition

Awards

Award of Merit

Iqra Private School2021

For taking second position in HSSC-2 exams

Ace of SPADES (2024)

SPADES (LUMS)

Awarded for excellence within the Society For Promotion Of Sciences And Engineering.

Patents & publications

Interpretation of emotions through finetuned BERT on diverse text-intensive datasets

research paper

With K. Vlastimil, S. Amir, C. Safee

Read it

Emotions and Sadness, Anxiety, Depression (S.A.D) detection using a multi-modal ML integration in Marginalized population

research paper

With C. Safee

Q&A

In Omer’s own words.

Why do you prefer "blank canvas" environments?

I thrive on "blank canvas" projects. Whether it's co-founding a startup or establishing a new AI department, I love the challenge of building teams, systems, and solutions from the ground up.

Why do you prefer 'blank canvas' environments?

I thrive in 'blank canvas' situations because they allow for maximum creative problem-solving and the ability to build systems correctly from the ground up. Whether it's establishing an AI department or co-founding a startup, I enjoy the challenge of turning a vision into a functional, high-impact reality.

Social impact & activities

STEM Officer

Society For Promotion Of Sciences And Engineering (SPADES)organizer

Led a 56-member team for Pakistan’s largest science olympiad. Managed CEO panels, secured sponsors, and spearheaded the country’s only student-run space observatory.

Impact

  • Secured major corporate sponsorships by pitching technical value and brand alignment.

Director Design

PhotoLUMSorganizer

Managed a team of 9; designed content for social media reaching 280,000+ followers.

Impact

  • Reached 280,000+ followers through social media content.

What drives the work

Motivations

  • Driving business growth through technical innovation
  • Solving complex client problems with AI
  • Bridging the gap between engineering and sales

Objectives

  • Transition into a Technical Sales or Solutions Engineering role within the AI/Biotech sector.
  • Leverage deep technical expertise to bridge the gap between complex engineering and business needs.

Principles

  • Innovation
  • Efficiency
  • Leadership
  • Interdisciplinary Problem Solving
Hire OmerReach out about a role, a contract or a conversation.For recruiters

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

More people on Kanvis

Browse more peopleCreate your own Kanvis