About

Owais Barkati is an engineer based in Mumbai, Maharashtra,India. He works at Multi Commodity Exchange of India (MCX) as a Graduate Engineer Trainee, and builds across two areas that rarely appear in the same person: multi-agent AI and retrieval systems, and the high-throughput data infrastructure those systems depend on.

The intersection is the point

Most engineers working on large language model applications have never run a Kafka consumer group in anger, and most engineers who have never fine-tuned a model. Owais Barkati has done both, and the combination changes how he approaches AI systems: an agent framework is a distributed system that happens to call a model, and it fails in distributed-systems ways — partial failures, retries that duplicate work, queues that silently back up, state that diverges between components.

That perspective shows up concretely in the work. The multi-agent system on this site routes queries to specialists rather than handing one agent every tool, because attributing a failure matters more than flexibility. The broadcast system moved fan-out off the database and onto an event log, because the problem was routing wearing a persistence costume. In both cases the interesting decision was about system shape, not about which library to use.

Research background

Owais Barkati is a published researcher in computer vision and natural language processing, with two peer-reviewed papers under the name Owais Ansari. The first proposes an interpretable instance segmentation framework for camouflaged wildlife detection, combining YOLOv8n-seg with Seg-Grad-CAM and LIME, and won Best Presentation at ADCIS 2025. The second, published by Springer at ICT4SD 2025, presents a multilingual video dubbing system supporting ten Indian languages.

The thread connecting them to his engineering work is interpretability. A segmentation mask you cannot explain and an agent answer you cannot trace are the same problem in different clothes — a system producing confident output with no account of how it got there.

What he is looking for

Owais Barkati is interested in roles and collaborations in AI and machine learning engineering, backend and distributed systems, and data engineering — particularly work where those overlap: production LLM and agent systems, retrieval infrastructure, and real-time data platforms. He is also open to research collaborations and selected freelance or consulting work. The quickest way to reach him is owess.dev@gmail.com, or the contact page.