Naveen Manikandan

Most AI engineers can't read a stack trace. Most backend engineers don't know what BLIP is. I sit in the gap between them. Recently I built a system to reduce hallucinations in vision-language model outputs. No retraining. Just smarter candidate selection using BLIP and CLIP. What I found surprised me that CLIP alignment alone made hallucination worse, not better. The fix came from understanding why the model failed, not just running more experiments. That thinking carried an 82% reduction in hallucination rate across 5,000 COCO images. Before that I spent 3 years at Tata Consultancy Services building production Java systems across microservices, ETL pipelines moving data across SAP, Oracle, GCP, Salesforce and Kronos, billing APIs on a retail CRM serving one of Australia's largest supermarket chains. Real systems. Real scale. Real deadlines. I also won Star of the Month twice at TCS and an On the Spot Award for implementing a production change with zero business disruption. Completed my MS in Computer Science at SUNY Buffalo (May 2026)with coursework in deep learning, distributed systems and ML. I'm at my best on teams where the engineering and the intelligence layer are treated as one problem, not two separate concerns. Available for full-time Backend Engineer, Data Infrastructure and AI Engineer roles immediately.naveen.cspro@gmail.com