Software Engineer · San Francisco
Dhanya Jayagopal
Software Engineer focused on designing and operating multi-provider LLM platforms at scale. Expertise in reliability engineering, routing architecture, and enterprise AI deployment from early-stage prototypes to production systems.
Asana · San Francisco, CA
- Designed and implemented a typed LLM error management framework, replacing a 75% 'unknown' failure bucket with structured classification and unified observability; stabilized uptime at 98%+ for 5 consecutive months (↑3pt floor).
- Re-architected LLM routing into an extensible multi-provider framework, enabling rapid onboarding of new providers with significantly reduced integration overhead.
- Leveraged Bedrock as both a primary provider and cross-provider fallback, capturing up to $300K in AWS credits while preventing ~23K failed requests per quarter.
Languages
Python, TypeScript, JavaScript
Frameworks
React, Django, FastAPI, Next.js
APIs & Protocols
REST, GraphQL, OpenAI/Anthropic
Cloud & Infra
AWS (S3, IAM, Bedrock), Docker, CI/CD
LLM Providers
OpenAI, Anthropic, Google Gemini, AWS Bedrock
LLM Systems
Multi-provider routing, cost optimization, rate limiting, context window management, streaming, RAG, adversarial evaluation
Resume
Dhanya Jayagopal
dhanya.jayagopal@gmail.com · San Francisco, CA
M.S. in E.E & Computer Science
University of California, Berkeley · GPA 3.9
Helen Tomasini Scholar
B.A. in Computer Science, Minor in Education
University of California, Berkeley · GPA 3.75
Regents & Chancellor Scholar
Asana · San Francisco, CA
- Designed and implemented a typed LLM error management framework, replacing a 75% 'unknown' failure bucket with structured classification and unified observability; stabilized uptime at 98%+ for 5 consecutive months (↑3pt floor).
- Re-architected LLM routing into an extensible multi-provider framework, replacing brittle switch logic and enabling rapid onboarding of new providers (e.g., Gemini) with significantly reduced integration overhead.
- Leveraged Bedrock as both a primary provider and cross-provider fallback, capturing up to $300K in AWS credits while preventing ~23K failed requests per quarter through dynamic traffic controls.
- Diagnosed chunk-level streaming hangs that bypassed existing retry logic and implemented iterator-layer timeout and retry safeguards, recovering ~32K stalled requests in month one.
- Redesigned multi-tier rate limiting to isolate workloads and prevent cross-feature interference, reducing sustained concurrency limit errors by 20%.
- Architected cross-provider context window management using vendor tooling and in-house LLM summarization to prevent overflow failures and maintain consistent behavior across model providers.
- Led cross-functional enterprise AI launch for ~250K Amazon users, aligning 10+ stakeholders across infra, product, security, and GTM to deliver on time under strict security and geo constraints; established a reusable BYO-AI model.
Udemy · San Francisco, CA
- Migrated learner-facing pages to Next.js with server-side rendering, cutting page load times by 50% for users in India.
- Built active-learning features (labs, assessments, learning paths) with event-driven engagement tracking to increase session depth and retention.
- Led GraphQL migration for a brand new onboarding flow, replacing REST endpoints and integrating with the frontend to improve performance and enable richer personalization.
- Launched a Django Signals–based deep link pop-up system to detect first-time vs. returning users, triggering targeted Braze campaigns that boosted enterprise mobile app adoption.
- Served as interim Technical Program Manager for accessibility vendor transition, coordinating legal, finance, engineering, and product teams while driving company-wide A11Y remediation.
UIST '22 – ACM Symposium on User Interface Software and Technology
Exploring the Learnability of Program Synthesizers by Novice Programmers
First Author — analyzed specification types, synthesis invocation methods, and feedback mechanisms to define design constraints for scalable program synthesis systems.
https://doi.org/10.1145/3526113.3545659
Languages
Python, TypeScript, JavaScript
Frameworks
React, Django, FastAPI, Next.js
APIs & Protocols
REST, GraphQL, OpenAI/Anthropic
Cloud & Infra
AWS (S3, IAM, Bedrock), Docker, CI/CD
LLM Providers
OpenAI, Anthropic, Google Gemini, AWS Bedrock
LLM Systems
Multi-provider routing, cost optimization, rate limiting, context window management, streaming, RAG, adversarial evaluation