# Chetan Ankola

Senior Product Manager · Agent Platforms and Evaluation

San Francisco Bay Area · [Portfolio](https://cankolabuilds.com/) · [LinkedIn](https://www.linkedin.com/in/chetanankola/) · [GitHub](https://github.com/chetanankola)

Product manager building agent platforms and tools for model evaluation. Hands-on with retrieval, model routing, human review, Swift, Python and GCP.

## Experience
### Yahoo | Product Management and Engineering

June 2012–present · Sunnyvale, CA · Current role: Senior Product Manager, Yahoo Mail

Engineer first: front-end engineer on Yahoo Search (JavaScript, Node.js, React), then iOS engineer on Yahoo Mail, before moving into product management.
- Built [OCEAN](https://cankolabuilds.com/work/ocean) to turn questions about Yahoo Mail plans into answers with inspectable sources. Available on the web and in Slack and coding tools, it handled 291 question submissions from July through September 2026 after excluding known creator and test activity.
- Ran OCEAN's quality loop end to end: automated evals and postmortem tooling, thumbs-up/down and verbal feedback capture, nightly query-and-response review, and admin analytics.
- Designed DROP's drop.ai.chat()/stream() model APIs with Codex and Claude support, provider selection and fallback, and usage reporting by provider, model, tokens, and cost.
- Led [mobile Calendar](https://cankolabuilds.com/work/calendar) with engineering and ML partners. Adoption reached approximately 10% of weekly active users in 2025. A 2024 launch of LLM-based event extraction improved time-of-interest visibility by 23.9% and recorded 76% satisfaction.
- Led product work on Digest and Spaces and prototyped new ways to surface relevant messages in Yahoo Mail.
- Started early generative AI and Claude Code communities at Yahoo so colleagues could share working demos and coding practices.
### Earlier Yahoo engineering and product work
- Owned the Yahoo Search Assist frontend library consumed by multiple internal Yahoo properties; rebuilt the mobile search frontend in Node.js and React.
- Built [Yahoo Mail AR](https://cankolabuilds.com/work/ar-ads) experiences and tools that let creative teams author ads. The work supported $2M in AR ad sales across advertisers including Home Depot, Macy’s, Pottery Barn and HP.

## Selected Agent Systems
### DexRelay | Phone-to-Mac agent control plane (App Store)
- Built the control path for launching and steering Codex and Claude Code sessions from an iPhone — QR pairing, LAN-first WebSocket transport, Tailscale fallback, and a launchd-managed Mac runtime with health, repair, and wake controls.
### Career OS | Source-backed personal AI job-search platform
- Built a synced Mac and iOS system where a 50-fact, provenance-tracked knowledge base grounds AI-generated resumes and interview prep — retrieval quality and source attribution as product surface.
### Local AI Chat With My Docs | Private retrieval system
- Built a SwiftUI client and FastAPI retrieval backend with source citations, ChromaDB persistence, and Ollama-first local inference, exposed securely over Tailscale.

## Skills
Agent platforms: Multi-model routing, evals, provenance, human review, RAG, MCP connectors, prompt design
Engineering: Swift/iOS, Python, Node.js, React, GCP, REST, WebSocket
Product: 0→1 execution, metric design, experimentation, quality bars for LLM features

## Education
University of Southern California | M.S., Computer Science | 2010 — 2012
PES Institute of Technology | B.E., Information Science
