Open to opportunities

Parth Tulsyan

@parthtulsyan

Machine Learning Engineer at FastCode.AI who built a P&ID perception pipeline that processed 1,000+ drawings and halved processing time.

India

Message

At FastCode.AI, I built an end-to-end visual perception and graph-analysis pipeline for P&IDs and PFDs, processing 1,000+ drawings for a Fortune 500 client and halving processing time. I also trained an object detector for engineering schematics that improved detection accuracy from 0.80 to 0.95 versus template matching.

At Verisk Analytics, I developed pipelines to deploy and monitor MLflow models on AWS, and implemented Terraform infrastructure and alerting. My research at Marmot Lab, National University of Singapore, on social behavior in multi-agent pathfinding was published in Artificial Intelligence.

Experience

Work history, roles, and key accomplishments

Current

Machine Learning Engineer

Current

FastCode.AI

Aug 2025 - Present (1 year 3 months)

Designed and trained an object detector for engineering schematics, improving detection accuracy from 0.80 to 0.95. Built an end-to-end visual perception and graph-analysis pipeline for P&IDs/PFDs, processing 1,000+ drawings for a Fortune 500 client.

MAE RT DETR Qwen GLM OCR LangGraph RAG LoRAs SFT DPO RLHF Databricks LiDAR Stereo Imaging

Data Scientist

Verisk Analytics

Jul 2024 - Aug 2025 (1 year 1 month)

Developed pipelines to deploy and monitor MLflow models on AWS ECS/EKS in distributed production environments. Implemented Terraform infrastructure and CloudWatch/Lambda alerting, and deployed PostgreSQL HA and Redis backends on Kubernetes.

MLFlow AWS ECS Amazon EKS Langfuse Terraform Cloudwatch Lambda PostgreSQL Redis Kubernetes

Research Intern

Marmot Lab, National University of Singapore

Jan 2024 - Aug 2024 (7 months)

Researched decentralized multi-agent reinforcement learning, showing that learned Social Value Orientation helps resolve conflicts in robot pathfinding. Designed a Graph Transformer communication model to improve inter-agent coordination.

Multi Agent RL Graph Transformers Social Value Orientation Pathfinding