Senior Machine Learning Engineer with 10+ years of experience building and deploying scalable AI systems across computer vision, NLP, generative AI, and MLOps. I've led end-to-end ML projects from real-time edge inference to LLM-powered applications, with a focus on performance, reliability, and impact. My work spans startups and enterprises, including Mashgin, Gensyn, Voyage, and Amazon. Skilled in Python, Go, Kubernetes, TensorRT, DeepSpeed, and cloud infrastructure, I specialize in shipping production-grade AI that solves real-world problems.
Mashgin
June 2021 - Present
- Computer Vision & Generative AI · Improved SKU recognition accuracy to 98.5% for 10K+ products using EfficientNet-B7/ViT models · Deployed quantized models (TensorRT/ONNX Runtime) on NVIDIA Jetson Xavier; reduced inference latency 500ms → 120ms · Built 3D synthetic data pipeline (Blender + ComfyUI); generated 100K+ images/year, cutting labeling costs by $250K/year · Implemented NeRF asset generation with NerfStudio, boosting novel item recognition by 18% - MLOps & Infrastructure · Built Kubernetes CI/CD with ArgoCD/Flux, enabling hourly rollouts to 3,700+ edge devices · Designed Feast feature store w/ Redis cache; reduced real-time feature latency 200ms → <50ms · Managed AWS infra (EC2, SageMaker, S3) with Terraform; achieved 99.99% uptime - LLMs & RAG Systems · Built RAG system with Llama-2 (LoRA), LangChain, and Pinecone; reduced support ticket time by 65% · Developed Stable Diffusion workflows (ControlNet, DreamBooth) for on-demand marketing content · Automated product imagery via DALL-E API integration - Data & Microservices · Handled 15TB/day transaction data via PySpark + Delta Lake on Databricks · Streamed events using Apache Kafka + Avro schemas · Built Node.js microservices w/ WebSockets for live inventory updates · Designed HIPAA/GDPR-compliant pipelines for healthcare clients - Financial Systems · Engineered Go-based gRPC payment router handling 5M+ transactions/day @ <10ms latency · Implemented chaos testing (Gremlin) and performance profiling (pprof)
Gensyn
January 2020 - May 2023
· Designed Verde protocol for decentralized ML validation using Rust, Solidity, ZK-SNARKs (Circom) · Built RL Swarm framework using Ray, PyTorch, and Horovod to scale P2P RL to 10K+ GPUs · Fine-tuned LLaMA-1/2 with DeepSpeed ZeRO-3 and LoRA adapters · Developed RAG pipelines using Hugging Face, LlamaIndex, and Weaviate for 10M+ research papers · Built Go-based P2P layer (libp2p) and deployed IPFS-based model storage (5× faster than S3) · Automated synthetic training data via Stable Diffusion + BLIP-2
Voyage
January 2017 - March 2021
· Trained YOLOv3/v4 and Mask R-CNN; improved pedestrian detection mAP by 15% · Optimized LiDAR (PointNet++, CUDA); reduced perception latency by 30% · Managed 50TB+/month data via Spark pipelines on Databricks · Built Airflow DAGs (KubernetesPodOperator) for auto-retraining · Deployed TFLite models with <50ms inference on AV fleets · Extended CARLA/Deepdrive with custom UE4 plugins; built C++ kernels for real-time fusion
Startup X
January 2014 - December 2016
· Augmentus (Robotics): Developed ROS-based path planning and WebGL (Three.js) visualization · XpertFlow (Healthcare AI): Deployed ECG classifiers (TensorFlow, 94% AUC), HIPAA-compliant DB design · StaffAny (SaaS): Built scalable Node.js + React system; used RabbitMQ for async workflows
University of Southern California, September 2014
Mathematics And Computer Science
Tsinghua University, September 2012
Computer Science
Verified Machine Learning Engineer
8+ years of experience
Preferred commitment: Full Time
Take the next step and bring this top talent to your team
Hire Tianchang for your team