Biswanath KarScaling AI Products Through People, Process & Data
AI Product Operations·AI/ML Operations·Human-in-the-Loop·LLM Evaluation·AI Quality·Product Strategy
I build and scale the operational systems that enable AI products to deliver reliable, high-quality outcomes. With 20+ years across Yahoo and Flipkart, I partner with Product, Engineering and Data Science teams to operationalize AI at enterprise scale.
Operationalizing AI at enterprise scale.
Executive leader translating AI ambition into repeatable operational systems — where product, engineering, data science and human judgment converge.
Where strategy meets execution.
AI-Native Workflows
RLHF, prompt feedback loops, GenAI GTM, and LLM evaluation at production scale.
HiTL Frameworks
End-to-end Human-in-the-Loop systems for model training, annotation, and ground-truth validation.
Quality Governance
Custom rubrics, inter-rater agreement, gold sets, and 'Check the Checker' auditing.
Knowledge Operations
Attribute enrichment pipelines, metadata accuracy, and multilingual content strategy.
Automated Pipelines
n8n, LangChain, and Pydantic-driven systems for autonomous data processing.
Scaled Architecture
100+ vendor ecosystems, SLA frameworks, and multi-regional program delivery.
Two decades of leading AI & operations at scale.
A career built on turning ambition into repeatable, high-quality outcomes across global product organizations.
Independent Research & AI Upskilling
- Advancing expertise in LLM workflows and GenAI product operations.
- Hands-on solo project development and technical research.
Senior Manager II — Catalog Product Ops & AI Knowledge Systems
- Deployed 900+ CV ML models across Catalog QC, driving ₹14 Cr annual savings.
- Architected HiTL feedback frameworks, enabling 5+ GenAI GTM launches yearly.
- Scaled AI content ops to 7,000+ assets/month with 8–10 bps conversion uplift.
- Launched vernacular knowledge strategy across 12 Indic languages (+7% DAU share).
- Led 100+ vendor ecosystem with <8% defect rate at scale.
Manager — Search Quality Assurance Operations
- Improved search knowledge quality by 5–8% via semantic query analysis.
- Cut time-to-knowledge for new SKUs by ~35% with redesigned indexing workflows.
Manager — Editorial, Search & Platform Knowledge Ops
- Led multi-regional teams improving NDCG across APAC, EMEA, US.
- Governed quality across 10+ global product launches at <1% defect escape.
Team Leader — Editorial & Compliance Operations
- Led team of 15+ on editorial research with 98%+ accuracy and compliance.
Where AI operations create business impact.
Selected initiatives showing how the right blend of people, process and data turns AI capability into measurable outcomes.
HiTL Operations Dashboard
Leadership lacked a single, trusted view of AI quality, evaluation accuracy and reviewer throughput across production models.
Designed an executive analytics surface unifying LLM evaluation, RLHF feedback loops and reviewer productivity into one operating cadence.
Cut decision latency on AI quality issues, aligned Product · Eng · Ops on a shared KPI language, and made model regressions visible in hours instead of weeks.
Scaling AI isn't just about models.
It is about building the right combination of people, process, data, technology and operational excellence.
Empowered, cross-functional teams with clear ownership and shared outcomes.
Repeatable operating rhythms that turn AI experiments into durable systems.
Trusted, well-governed data as the foundation for every model decision.
Right-sized architecture matched to the business problem, not the trend.
Quality, cost and velocity treated as one integrated discipline.