AI Automation & Software Engineer
Applied AI, backends, distributed systems
Dubai, UAE · hadhirasal22@gmail.com
I build applied AI and automation into live operations — and the backends underneath them. Enterprise workflow automation, retrieval systems and internal platforms across the UAE and India, in Python, TypeScript and Java, with a standing interest in system design, distributed systems and database internals.
Available for work · Dubai, UAE
15,000+
Employees running on HR workflows I automated
Etihad Airways
100+
Workflows automated end to end
Automation
60%
Faster HR answers from the AI assistant I built
Etihad Airways
3+
Years shipping production systems
UAE · India
Production systems, automation delivered inside large organisations, and the things I built from scratch to understand the layer underneath. Open any project for the workflow, the build and the stack.
Exhibition CRM that turns floor plans into pipeline — reads PDFs, images and interactive plans from ExpoFP, Map Your Show and a2z, dedupes exhibitors across shows, finds the decision-makers and tracks outreach per show.
AI exhibition stand designer — turns a client brief and the brand's website into three editable 3D stand concepts at exact size, photoreal renders with the brand's logo, and an A3 drawing pack plus layered AutoCAD DXF a stand builder can quote from.
Approvals, international assignments, compensation, benefits and document workflows across Microsoft 365, SharePoint, SAP and Power BI — including payroll anomaly detection.
Policy-grounded assistant that cut average HR response time by over 60%, wired into ticketing, approvals, escalations and Teams notifications.
Fault-tolerant workers with visibility leases, exponential backoff, idempotency keys, dead-letter routing, heartbeat reclaim for crashed workers and queue-depth metrics.
Hierarchical Navigable Small World indexing built by hand: multilayer graphs, greedy descent, heuristic pruning, custom distance kernels, persistence and recall-vs-latency benchmarks.
Paged storage engine with slotted pages, a buffer pool, B+ tree indexes, write-ahead logging and crash recovery — then a parse → plan → execute query path on top.
Decoder-only language model with no high-level LLM libraries — BPE tokenizer, causal multi-head attention, scheduled training, then sampling with temperature, top-k and nucleus.
Catalogue SaaS for Instagram and WhatsApp sellers — upload products once, get a branded storefront link, and take enquiries on WhatsApp instead of re-sending the same photos in DMs.
Languages
Backend & systems
Data & AI
Frontend
Cloud & automation
GitHub
Location
Dubai, UAE