Abbas Kazmi

whoami

Abbas Kazmi

Software Engineer

👋 Hey, I'm Abbas! I'm a Software Engineer at Adobe, where I get to work across agentic search, backend infrastructure, and full-stack development to help ship AI-powered features people actually rely on. Before this I interned across Adobe's checkout and eCommerce systems, and had a blast serving as Technical Director for HackDavis while studying Computer Science at UC Davis with a focus on deep learning.

Outside of work, you'll usually find me on a soccer field or pickleball court, or catching up on movies. Always happy to chat, so feel free to reach out!

experience
Adobe — Software Engineer
[2024.08 — present]
New York, NY
  • Built an agentic search service and Adobe's first context-aware recommendation app, with agents selecting search facets from document and parent-app context, adopted across 380 customer accounts.
  • Led phased monthly releases of Adobe Experience Manager across 20K environments for 2,000 customers, owning rollback decisions, and built a triage agent that cut manual ticket review from 10 to 2 hours per release.
  • Rebuilt a legacy Java asset download API to redirect to short-lived Azure Blob SAS URLs, shipped behind feature flags, cutting p95 time-to-first-byte from 2.8s to 230ms across 300K monthly requests.
  • Engineered an OpenTelemetry tracing pipeline ingesting 7M agent traces monthly into Langfuse, then fine-tuned agent LLMs on 100K labeled traces, raising task success from 25% to 70%.
  • Developed an agent evaluation system that flags unstructured outputs and silent failures, cutting undetected failures 75% and surfacing issues whose fixes raised result relevance 45%.
  • Designed a security filter and alerting system that strips prompt injection and malicious content before it reaches the LLM, catching 20K+ risks monthly and cutting triage time 60%.
  • Created a session-based agent orchestration harness from scratch with built-in cost tracking and observability, adopted across 4 orgs to serve 500 customers and cutting agent setup from days to minutes.
Adobe — Software Engineer Intern, Machine Learning
[2023.06 — 2023.09]
San Jose, CA
  • Spearheaded development of machine learning models to predict user checkouts by leveraging high-frequency clickstream data.
  • Led end-to-end pipeline development on Databricks using SQL and Python to transform raw data into actionable insights.
  • Achieved ~93% accuracy models through various machine learning methods, integrated via meticulous software engineering.
  • Planned model deployments and microservice integrations to enable real-time predictions on Adobe.com.
Adobe — Software Engineer Intern, eCommerce Experiences
[2022.06 — 2022.09]
San Jose, CA
  • Developed a testing framework for the Adobe Checkout team to validate new, existing, and error scenarios using mock data.
  • Restructured the checkout application's distributed system design by incorporating AWS S3 and EC2 to store and return mock responses.
  • Designed a testing API in TypeScript with Apollo GraphQL that integrates within existing APIs for seamless testing.
  • Implemented Redis caching through an ioredis server for fast retrieval of repeated calls and log analytics.
HackDavis — Technical Director
[2022.09 — 2024.09]
Davis, CA
  • Served as Technical Director for HackDavis, one of the largest collegiate hackathons in California, where over 950 students, creators, and leaders come together to build for social good.
education
University of California, Davis
B.S. Computer Science and Engineering — Concentration in Deep Learning & AI
Deep Learning, Computer Vision, Systems Programming & Operating Systems, Computer Architecture, Algorithm Design & Analysis, Software Engineering
stack
agent / AI frameworks
LangChainLangGraphCrewAI LangFusePyTorchAWS Bedrock Azure OpenAI
languages
PythonJavaTypeScript C++Go
tools / platforms
LinuxDockerSQL MongoDB AtlasGraphQLApache Spark React.jsAWSAzure