Profile
Machine Learning Research Engineer with 8+ years of experience building AI systems for industrial inspection, federated learning, and computer vision. Currently at Mindtrace (Manchester), building distributed monitoring agents for weld inspection workflows and leading 3D deep learning research. Published at IEEE and Springer; work under review at IEEE Transactions on Neural Networks and Learning Systems.
Experience
Manchester, UK
- Designing the agents package for the Mindtrace open-source framework — autonomous monitoring agents for distributed weld inspection workflows, coordinating tasks and maintaining memory across sessions.
- Designed monorepo architecture and built the core package: config injection, structured logging, and observable state primitives.
- Self-supervised learning with DINO architecture features on large-scale unlabelled datasets; parameter-efficient fine-tuning via LoRA achieving 95% accuracy on industrial inspection tasks.
- 3D deep learning for point cloud segmentation: trained and benchmarked PointNet, PointNet++, and SPVCNN on LiDAR-captured infrastructure data.
- Built LiDAR preprocessing pipelines (PDAL, GeoPandas, Shapely) for coordinate transformations, ground filtering, and tiling at scale.
- Deployed models on AWS, GCP, and Azure; MLOps infrastructure with Docker, MLFlow, and FastAPI microservices (92% detection accuracy).
India
- Deployed YOLOv3 on the Kneron KL520 USB AI dongle — model quantisation, ONNX export, and live video inference pipeline.
- OCR pipeline using Google Vision API and Tesseract; object identification with FLANN/SIFT.
- Serverless architecture on AWS (Lambda, API Gateway, S3).
India
- Built a computer vision system for classifying life science documents — text analysis, spatial layout detection, and image analysis using Faster RCNN and Mask RCNN.
- Developed ML deployment infrastructure: Docker + Kubernetes + Helm templates with FastAPI microservices and GitLab CI/CD pipelines.
- Classification tracked using F1, Precision, and Recall.
India
- Contributed to CI/CD pipelines for banking applications using Jenkins and TeamCity, with automated post-deployment testing.
India
- Worked in Banking and Finance domain with Bank of America, Royal Bank of Canada, and ING Bank as clients.
- Provided onsite L1/L2/L3 production support for Anvil Classic in Amsterdam for ING Bank.
- Managed application incident resolution (P1/P2/P3) and coordinated support change release management.
- Analysed downstream trading systems — BOTS, Murex, Calypso, ACR — using Tivoli and Sentinel monitoring tools.
Education
Dissertation: Federated self-learning chatbot with privacy-preserving dialogue adaptation. Research spans federated learning, NLP, computer vision, and medical image segmentation (HuBMAP Kaggle, top 14%). Built autonomous robot navigation systems using ROS2 — LIDAR sensing, PID and fuzzy logic control, state-machine-driven maze traversal.
Publications
MP-CFL: Matrix-Profile Clustered Federated Learning for Communication-Efficient Training under Data Heterogeneity
IEEE Transactions on Neural Networks and Learning Systems (TNNLS)
· Under Review, 2025
Evaluating Dialogue Adaptability: A Comparative Study of Self-Feeding Mechanisms in Federated and Centralized Chatbot Architectures
Workshop on Secure AI Applications (SAIA 2025) · IEEE
Privacy-Preserving Self-Learning Chatbot with Federated Intelligence
International Conference on IoT, Electronics and Mechatronics (IEMTRONICS 2025) · Springer
Technical Skills
Languages
Python, C++, SQL
ML Frameworks
PyTorch, TensorFlow, ONNX, Scikit-learn
Computer Vision
DINO architectures, YOLO, DeepLab, SAM 2, LoRA, PointNet, PointNet++, SPVCNN
Federated Learning
FedAvg, FedProx, TensorFlow Lite (on-device), MQTT, RabbitMQ
Cloud
AWS, Google Cloud Platform (GCP), Microsoft Azure
MLOps
Docker, Kubernetes, Helm, MLFlow, FastAPI, GitLab CI, Jenkins
Data / 3D
PDAL, GeoPandas, Shapely, NumPy, Pandas
Achievements & Certifications
- Oct 2022Google Developer Badge — Machine Learning Bootcamp
- Oct 2022TensorFlow Developer Certification
- Sep 2022HuBMAP Organ Segmentation Kaggle Competition — Top 14% globally
- Aug 2022Coursera Deep Learning Specialization
- 2018–2019Post Graduate Program in Artificial Intelligence and Machine Learning
Open-Source Contributions
Mindtrace — Open-Source AI Infrastructure Framework (
github.com/Mindtrace/mindtrace)
Contributing to modular micro-services for ML orchestration, observability, and agent-based automation.
MP-CFL — Matrix-Profile Clustered Federated Learning (
github.com/panks11/mp-cfl)
Implementation of the MP-CFL framework — event-driven clustered federated learning with matrix-profile change detection.
Federated Chatbot — Privacy-Preserving Self-Learning Chatbot (
github.com/panks11/federated-chatbot)
Companion code for the IEMTRONICS 2025 and SAIA 2025 publications on federated dialogue learning.