AI Engineer · Backend & Software Engineer
I build backend systems for electronic invoicing and do machine learning research on language, speech, brain signals and medical images. I am a computer science graduate and first author of two peer-reviewed papers.
Looking for PhD research positions and Master's programmes in machine learning.
Industry work and research that fill most of my time right now.
Since July 2026 I build and support electronic invoicing integrations for enterprise clients in the UAE, Saudi Arabia and Oman, connecting ERP systems such as SAP S/4HANA, Tally, Zoho and QuickBooks to the platform.
As a part-time Junior Research Fellow at Thapar on an Xebia-sponsored project, I test how vision-language models respond to prompt injection and jailbreak attacks delivered through images and text.
A pipeline that writes chest X-ray reports from MIMIC-CXR images and MIMIC-IV clinical data. No patient appears in both the training and test sets, and generated reports are scored for factual consistency. Work in progress.
Two first-author papers, both peer reviewed.
Multilingual summarisers often write fluent summaries that drift from what the article says. We fine-tune mT5-Large on XL-Sum in five languages with an alignment loss that ties each article to its translation, then rerank candidate summaries by factual entailment (mDeBERTa-v3) and semantic centrality (MPNet).
Read speech sounds flat in a dialogue. PACC adjusts the intonation, stress and rhythm of read speech with a deep neural network and resynthesises it with a HiFi-GAN vocoder. The work came out of the Samsung PRISM programme, where we recorded 100 read and 100 conversational speech samples for training.
From a medical imaging pipeline in progress to deployed accessibility tools.
A pipeline that generates chest radiograph reports from MIMIC-CXR images, using clinical information from MIMIC-IV.
An urban heat planning tool built by a four-person international team, with a validation partner in Accra, Ghana. I was the technical lead.
An end-to-end EEG pipeline on two public datasets (MUSIN-G and NMED-RP), treating music perception as both classification (emotion and rhythm) and regression (how much a listener enjoyed a song).
An accessibility tool that describes an image with BLIP, translates the description with Google Translate and reads it aloud with gTTS, in more than 20 languages.
Backend engineering, research, presales and internships.
If you supervise work on multilingual NLP, speech, multimodal safety or medical AI, or you have an opening, I would be glad to hear from you.
AI answers can be wrong. Messages are sent to Google's Gemini through a small server, so please don't share personal information.