Pankhuri Kulshrestha
Applied AI and ML engineer based in Manchester who enjoys taking a technically ambiguous problem through to a usable system. I’m currently a Machine Learning Research Engineer at Mindtrace, building computer-vision and agent-based systems for industrial asset inspection — spanning data and model development, deployment infrastructure, and communicating technical trade-offs to stakeholders. I’m seeking Forward Deployed Engineer opportunities where I can partner closely with users to ship dependable AI-enabled solutions.
What I Build
- AI systems for real workflows: translate inspection and operational problems into practical ML and automation solutions.
- Production ML foundations: build data preprocessing, model-serving, containerisation, and cloud deployment workflows.
- Computer vision and spatial AI: deliver 2D and LiDAR-based perception for asset inspection.
- Agent-based automation: develop distributed monitoring agents that coordinate work, preserve context, and surface anomalies to operators.
Selected Impact
| Mindtrace | Build ML and agent-based capabilities for industrial inspection, spanning LiDAR preprocessing, vision models, cloud deployment, and operator-facing workflow automation. |
| Model delivery | Fine-tuned DINO models with LoRA to 95% accuracy on a target industrial inspection dataset while reducing annotation needs through self-supervised learning. |
| ML platform | Created reusable Docker, Kubernetes, and Helm templates for serving models as microservices with inference APIs and CI/CD hooks. |
| Applied research | Published two peer-reviewed papers on privacy-preserving federated learning for conversational AI. |
Selected Work
Mindtrace — Open-Source ML Infrastructure Framework Contributor · Core Package & Agents Building reusable infrastructure and distributed monitoring agents for weld-inspection workflows: coordinating tasks, preserving operational context across sessions, and surfacing anomalies to operators.
ML Deployment Infrastructure Production MLOps · IQVIA Reusable Docker, Kubernetes, and Helm templates for packaging and serving ML models as microservices, including persistent-volume mounting, FastAPI inference scaffolding, and CI/CD hooks for repeatable delivery.
Privacy-Preserving Self-Learning Chatbot with Federated Intelligence IEMTRONICS 2025 · Springer Federated learning framework enabling chatbots to learn from user interactions without centralising private data — keeping sensitive conversations on-device throughout training.
Evaluating Dialogue Adaptability: Self-Feeding Mechanisms in Federated and Centralized Chatbot Architectures SAIA 2025 · IEEE Comparative study of federated vs. centralised training dynamics — analysing convergence, adaptability, and where the privacy-performance trade-off bites.
Autonomous Robot Navigation with ROS2 MSc Coursework · University of Essex ROS2-based autonomous navigation system using LIDAR sensing, a 5-stage state machine, and dual control strategies — PID for wall-following and fuzzy logic for gap navigation — benchmarked through a maze environment.
Technical Focus
- Customer-facing AI delivery and technical discovery
- Production ML, cloud deployment, and MLOps
- Computer vision & 3D deep learning (LiDAR, PointNet, DINO architectures)
- Distributed agents, memory, and workflow automation
- Federated learning & privacy-preserving ML