Machine Learning Engineer - Splunk - Cisco - San Jose, CA
(2025-10)
- Design and build scalable foundational components with orchestrator, memory, and llm service components to support product wide AI Agentic development; onboarded 10+ teams to integrate with built components to achieve more transparent AI Assistant answering and reduced LLM hallucination.
- Proposed and fine-tuned intent detection model on sentence transformer using semantic gap separation strategy to identify domain-specific intent from natural language with over 90% F1 score.
Machine Learning Engineer - AppDynamics - Cisco - San Jose, CA
(2022-08 - 2025-10)
- Designed and developed agentic system to facilitate troubleshooting an application performance monitoring (APM) platform; enabling dynamic decision making with LLM support when navigating otherwise complex troubleshooting flows. Largely reducing Mean-Time-To-Investigate (MTTI) (from hrs to < 30s).
- Developed and deployed RAG system in production to help customers gain in-depth visibility of the behavior, health, and performance of their application, services, network, and storage with natural language, using Information Extraction, VectorDB and LLMs.
- Reduced VectorDB document retrieval speed by 40% and improved top-1 accuracy of relevant documents by more than 4 times through tuning retrieval algorithm; improving overall performance for natural language driven data insights and troubleshooting.
- Designed and built root cause analysis service to efficiently identify microservice abnormalities by analyzing service interaction patterns and OpenTelemetry distributed traces.
- Developed and deployed log analysis service for SaaS systems to help users and SREs reduce time to triage (TTT) and mean time to repair (MTTR) by identifying log patterns, change points, and anomalies of logs from performance critical services.
Software Engineer, AI Platforms - Black Sesame Technologies Inc. - San Jose, CA
(2021-04 - 2022-08)
- Led development of enhanced quant-aware-training framework on top of PyTorch to boost computer vision model performances on ML accelerating chip. Accelerated models include ResNet, 3-D Lidar model, RetinaNet, and Detectron2 models.
- Developed quantization evaluation framework to benchmark quantization algorithm performances on image classification and object detection models such as MobileNetV2, CenterNet, and YoloV3 on Python with Pandas and Seaborn.
- Developed Tensorflow to ONNX model conversion tool for quantized Tensorflow model using Python.
- Worked closely with Jupyter Notebook in docker environment to prepare model zoo examples for clients.
Software Engineer - Cornami Inc. - Campbell, CA, US
(2019-05 - 2021-03)
Project: Design and build AI Framework for high-performance AI accelerator Chip; achieving high speedup for ML operators.
