Open to international AI opportunities
AI Solutions Engineer | Enterprise AI & Customer Delivery
I help enterprise teams design, evaluate, and deliver AI systems. My work connects production voice AI and cloud delivery with hands-on retrieval, agent evaluation, and voice-enabled robotics—grounding technical decisions in customer needs and evidence.
01 / Customer delivery
Customer-facing solution design, integration, and production delivery at ZEGO.
Customer delivery · ZEGO
Led solution design and customer delivery for a production voice AI companion. Aligned customer requirements, integration testing, and engineering work across the real-time audio and speech-to-response pipeline.
02 / Independent work
Self-directed projects in voice agents, robotics, search, and model evaluation, alongside an independent strategy exercise.
One project in two parts: a real-device voice agent and a robotics prototype. Future exploration includes SLAM, OpenVINS, and LLM + robotics.
Personal prototype
Connected real-time conversation to a physical robot’s expressions, with MCP tools and device completion receipts. A personal prototype exploring reliable AI-to-device interaction.
Personal experiment
An exploratory robotics project combining Gazebo simulation, ROS 2 sensor recording, and initial localization tests. A foundation for further experiments with Doly.
Two studies: selecting a search pipeline, then testing whether its answers stay supported by evidence.
Personal experiment
Built a repeatable three-pipeline comparison. You.com + Qwen led judged correctness on the 50-question English sample; GLM Quark had the faster external-search median. The result exposed a quality–latency trade-off for further testing.
Personal experiment
Extended the search framework to distinguish evidence-grounding risk from useful answers. Evaluated 120 bilingual adversarial cases across three pipelines, with 360 saved rows and 31 tests across the two evaluation suites.
Three studies: retrieval coverage, fine-tuning with fixed evidence, and a 330-question local RAG versus remote API comparison.
Personal experiment
Selected a heuristic GraphRAG profile after comparing ten retrieval configurations on 330 cases. It recovered 821/1,553 expected sources (52.9%), yet covered every required source for only 22/313 source-bearing questions (7.0%).
Personal experiment
Used paired evaluation to diagnose a severe fine-tuning regression and guide four QLoRA iterations. On the 330-case benchmark, V4 scored 85.02 versus 84.73 for its Base control: a small +0.29-point gain that still needs replication and expert review.
Personal experiment
Expanded a pilot to 330 tax-review questions. Base Qwen3-4B + RAG averaged 75.33/100 against 81.52 for Qwen3.7 Plus without RAG. The local system led on a 16-question evidence-gap subset, supporting targeted evaluation rather than a full replacement.
Personal experiment
Compared four agent workflows on one proposal task: more agents did not reliably preserve better evidence. Separately documented 19 product issues for engineering triage, including 12 involving coordination or context isolation.
Personal prototype
Built a research-assistant prototype connecting You.com web retrieval with GLM analysis. The deliverable is a search-to-answer integration and a documented workflow for structured comparisons.
Independent strategy exercise
Produced an account-strategy presentation with stakeholder hypotheses, a workload-based search-routing proposal, and a staged pilot plan. Connected technical evaluation to discovery, acceptance criteria, and account operations.
03 / Professional experience
Aug 2026 – Present · OpenCSG
Work on enterprise AI solutions and CSGClaw product validation, connecting platform behavior with practical usage requirements.
Tested multi-room agent workflows, context isolation, artifact sharing, and startup behavior. Authored 19 reproducible issue reports for engineering triage, including 12 coordination or context-isolation findings; all were open in the September 4 export.
Jul 2025 – Nov 2025 · China Telecom Global
Managed Alibaba Cloud account coordination for overseas data-center operations and cross-border connectivity, connecting customer requirements with regional partners and delivery teams.
Aligned customer and supplier teams on service scope, partner responsibilities, delivery readiness, and acceptance planning for overseas data-center operations.
Mapped APAC and EMEA bid requirements to regional partner coverage, dedicated connectivity, data-center operations, and managed-service staffing.
Maintained account-operation follow-up across delivery, acceptance, reconciliation, invoicing, and payment, keeping outstanding customer, partner, and finance actions visible.
Sep 2021 – Jul 2025 · ZEGO Technology
Led technical discovery, architecture reviews, and proof-of-concept work for real-time communication, messaging, and voice AI. Connected customer requirements with product, engineering, QA, and delivery.
Led solution design and customer delivery for a production voice AI companion. Aligned requirements, API compatibility, integration testing, and latency work across the audio, speech, and model pipeline.
Designed a repeatable RTC evaluation covering media quality, constrained bandwidth, noise, and network handover. Documented test methods and observations to support SDK integration and technology selection; the recorded launch timing was still undecided.
Provided solution support for cloud-game experiences in the Huiwan / WePlay live and voice-room portfolio. The available record establishes support scope, without a separately verified launch or usage result.
Jul 2020 – Sep 2021 · China Mobile Suzhou Research Institute
Served as the technical contact for enterprise cloud customers, aligning solution validation, partners, and post-sales escalations.
Led a RMB 15M hybrid-cloud education program from discovery and proof of concept through tender, implementation, and acceptance. Coordinated cloud, connectivity, and software/hardware partners, then supported application migration and production troubleshooting. The amount describes program scale.
Jul 2018 – Jun 2020 · China Telecom Zhejiang
Developed enterprise cloud solutions, partner pricing, and technical enablement, with delivery work in security services and container platforms.
Introduced a DBAPPSecurity endpoint-security offering into the cloud service catalog by coordinating deployment testing, partner pricing, product review, and contracts through launch.
Turned Kubernetes, Ingress, and cluster-scaling issues into product feedback and coordinated integration testing with Hyperchain / SUPCON and engineering teams.
Aug 2015 – Aug 2016 · Wacai
Built image-processing and data-preparation tools for computer-vision experiments and application-log analysis.
Built CAPTCHA annotation and recognition workflows using segmentation and support-vector methods, and implemented barcode localization in C++ and OpenCV.
Collected and cleaned datasets with Python, Pandas, and Shell; used Hive SQL to investigate application execution time in logs.
04 / Capabilities
Separate retrieval failures from answer quality, and compare changes against a controlled baseline.
Connect APIs and SDKs, reproduce failures, and turn findings into actionable technical reports.
Translate workload constraints into architecture, validation plans, and partner responsibilities.
Clarify requirements, align acceptance criteria, and keep delivery and escalation owners connected.