Timeline.

Sentiment AI

AI Engineer / Technical Lead

Own the full technical stack for an enterprise AI behavioral-analytics product in production, from LLM architecture and agentic pipelines to inference infrastructure and deployment.

• Designed a multi-tenant pipeline for long-context inference at approximately 55K tokens per call, with a node-based DAG executor on Temporal and a Postgres semantic layer.
• Benchmarked open-weight models on dedicated 2×H100 endpoints against hosted APIs for cost, latency, and quality.
• Built tooling with PydanticAI, DSPy, LiteLLM, RAGAS, and Langfuse, plus a framework that runs up to 40 customizable voice agents and scores their conversations with an external LLM judge.
• Engineered emotion-aware conversational behavior and real-time speech-to-text, nudge, and analytics integrations using Genesys AudioHook and Dynamics 365.
• Led regional inference, data-residency, and compliance architecture, including SOC 2 requirements, for regulated enterprise and government-adjacent deployments.

14th EAI ArtsIT Conference · Dubai, UAE

PromptRefine

A framework that refines vague, ambiguous, or biased prompts before they reach LLMs. Evaluated across OpenAI and Anthropic models for clarity, intent alignment, and bias mitigation.

Heriot-Watt University

BSc Computer Science (Honours)

First Class Honours, 4.0 GPA. UAE Golden Visa recipient.

AI-powered research assistant

PaperBot

Built an end-to-end research-paper QA system with LangChain, FAISS, GPT, and a Django REST backend. Cut query response time by 50%.

ASL / How2Sign

Continuous Sign Language Recognition

Built a ResNet18 + BiLSTM system for continuous sign recognition from video, with optimized preprocessing and feature extraction.

5th International Conference on Computers and AI Technology (CAIT 2024) · Hangzhou, China

To BI or not To BI?

An evaluation of deep learning and transformer models for suicidal ideation detection, accepted at CAIT 2024. BERT achieved 98.6% accuracy in the study, outperforming the LSTM-CNN and Bi-LSTM-CNN baselines evaluated.

Chalmer St. Consulting

Software Engineer Intern

Built a scalable auditing platform with React.js, Node.js, and Firebase.

• Refactored data models and real-time pipelines for low latency, improving operational productivity by 30% and supporting production use across multiple teams.