Headline

AI Product Manager | Computer Vision & Edge AI | Cut False-Alert Rates 30% & Response Time 18%

About

I build AI products that work in the real world — not just in demos. With 3 years owning the full lifecycle of production computer vision systems, I sit at the intersection of deep ML understanding and structured product delivery, translating model trade-offs into roadmap decisions that stakeholders can fund and engineers can ship. My most meaningful work has been at Graymatics, where I led a YOLOv8 occupancy-monitoring product from discovery through a 6-month post-launch KPI cycle across three enterprise sites. By authoring a rigorous model evaluation framework — precision-recall benchmarks, <40 ms latency thresholds, 99.9% uptime SLAs — and driving TensorRT inference optimisation, I improved safety-detection accuracy by 71% on held-out validation data and cut false-alert rates by 30%. That work taught me that good AI product management isn't about the model; it's about defining the right acceptance criteria before a single training run. I've also shipped a 0-to-1 traffic-incident analytics platform on a government SLA, leading discovery with 15+ Smart City officials, managing a 40-item Jira backlog using RICE and MoSCoW, and delivering with zero critical post-launch defects. A 90-day A/B notification experiment I designed cut manual review effort by 60% — proving that structured experimentation belongs inside the AI PM toolkit just as much as model evaluation does. Before moving into product, I was a software engineer building Python deployment-automation frameworks and shaving 56% off GitLab CI/CD pipeline runtimes. That engineering background means I write REST API specs and data contracts that developers actually trust, and I can pressure-test an MLOps architecture in a design review — not just nod along. I'm currently targeting Senior AI PM or Technical PM roles in the EU/UK, where I can bring both the depth to debate model architecture choices and the rigour to ship reliably inside complex enterprise environments. Open to relocation. Actively studying German (targeting B1 by Q4 2026)

Location

Bengaluru

AI stack

aistudio

claude code

claude

deepseek

gemini

ollama

Work experience

Ai Product Manager at Graymatics, Mar 2023 - Mar 2025 · 2 y, Banglore, Graymatics is a computer vision and AI analytics company building Smart City and enterprise surveillance intelligence platforms. In this role I owned the complete product lifecycle — from stakeholder discovery and PRD authorship through Agile delivery, model evaluation, and post-launch KPI monitoring — leading a cross-functional squad of 6 across engineering, QA, and compliance. Drove end-to-end delivery of a YOLOv8 occupancy-monitoring system across 3 enterprise sites, cutting false-alert rates by 30% and improving AI detection reliability by 12% — achieved by authoring a precision-recall model evaluation framework with strict false-positive thresholds and a <40 ms latency SLA. Launched a 0-to-1 traffic-incident analytics platform on a government contract with zero critical post-launch defects across 3 sprints, leading discovery with 15+ Smart City officials and managing a 40-item Jira backlog using RICE and MoSCoW prioritisation to keep delivery on schedule. Designed and shipped a 90-day A/B notification experiment — defining acceptance criteria upfront and monitoring results in Mixpanel — that reduced manual review effort by 60% and cut operator response time by 18%, with the winning variant reaching production within the same sprint cycle. Accelerated squad velocity by 25% over two quarters by authoring REST API specs and data contracts that compressed regression cycles from 3 days to under 1, and secured executive roadmap sign-off in the first review session for two consecutive quarters by pairing responsible AI constraints with quantified business cases.

Software Engineer at Graymatics, Nov 2022 - Mar 2023 · 5 mos, Bengaluru, Karnataka, India, Started at Graymatics on the core engineering team, focused on deployment infrastructure and CI/CD performance for production AI systems. This role built the hands-on MLOps and pipeline engineering foundation I later applied directly to product decisions as an AI PM. Architected a Python deployment-automation framework that standardised environment provisioning across 4 production rollouts, cutting onboarding time by 50% — from 20 hours to under 10 — and eliminating environment-mismatch failures that had caused recurring deployment delays. Slashed GitLab CI/CD pipeline runtime by 56% (45 min → under 20 min) by parallelising jobs, removing redundant test stages, and introducing artifact caching — recovering approximately 2 hours of engineering idle time per sprint and compressing the feedback loop for the entire development team.

Education

b.tech, cse, CMR University