Senior AI Engineer & Applied Scientist
Agentic & Generative AI · RAG · LLMOps · Multi-Agent Systems · Computer Vision · MLOps
LinkedIn · Publications · Email · 📍 Ontario, Canada
I turn frontier AI research into production systems. Over 14+ years I've taken ideas from a whiteboard to deployment across computer vision, NLP, and multi-modal intelligence — for energy, industrial, healthcare, and defense applications. Today I build agentic and generative AI: RAG assistants, multi-agent orchestration, and LLMOps that move enterprise metrics.
My foundation is research: a PhD in Machine Learning, a US patent in geospatial AI, 15+ peer-reviewed publications (175+ citations), and federally-funded R&D with DARPA, NGA, DoD, DoE, and the EU.
Currently Senior AI Engineer at Tiger Analytics, architecting GenAI systems for Fortune 500 clients.
| Area | Focus |
|---|---|
| 🤖 Generative & Agentic AI | RAG pipelines, multi-agent orchestration, LLM fine-tuning (QLoRA), prompt & context engineering, evaluation frameworks |
| 👁️ Computer Vision & Multi-Modal | Object detection & segmentation, zero-shot vision, photogrammetry, audio-visual-text fusion, video analytics |
| ⚙️ MLOps & Edge AI | CI/CD for ML, model monitoring, microservices, on-device inference (quantization, pruning, distillation) |
| 🔬 Research & R&D | Media forensics, medical imaging, geospatial intelligence, self/semi-supervised learning |
- 5× field-engineer productivity — built a RAG conversational assistant (LangChain · LangGraph · AWS Bedrock · FAISS) over technical manuals & maintenance data for enterprise infrastructure clients.
- 67% inspection-cost reduction ($15→$5/pole across 8 states) — pioneered a zero-shot detection pipeline (YOLO-World + SAM) with RANSAC photogrammetry, no domain fine-tuning required.
- 10× support efficiency (30 min → 3 min/case) — architected a company's first GenAI products (Vertex AI · Gemini) for video/audio-to-text summarization.
- DeepQC — automatic MRI quality assessment with deep CNNs reaching 89% agreement with expert radiologists, cutting QC time from 5 min to 30 s/scan (SPIE Medical Imaging 2020).
- US Patent 10,101,466 B2 — GPS-denied geo-localization combining SLAM, Structure-from-Motion, and GIS digital twins.
- canaid — multi-agent contact-center harness on AWS Bedrock (LangGraph · multi-vendor models · RAGAS/TruLens/LangFuse · CDK)
- agentic-predictive-maintenance — a Bedrock maintenance agent + ML health model with human-in-the-loop and a live React dashboard
- gas-and-energy-mechanics-copilot — RAG copilot for energy-engineering docs (Strands Agents · Bedrock Nova · FAISS)
- synthetic-data-gen-eval — a reproducible benchmark for tabular synthetic data, scored on fidelity · utility · privacy
- utility-pole-ai-photogrammetry — zero-shot measurement from field video (YOLO-World + EfficientSAM)
GenAI / LLM · LangChain · LangGraph · CrewAI · LlamaIndex · DSPy · RAG · AWS Bedrock · Vertex AI (Gemini) · Hugging Face · QLoRA
Vision / DL · PyTorch · TensorFlow · Keras · YOLO / YOLO-World · SAM · Mask R-CNN · U-Net · GANs · OpenCV
MLOps / Cloud · AWS (SageMaker · Lambda · EKS · S3 · ECR) · GCP (Vertex AI) · Docker · Kubernetes · Airflow · TFLite · ONNX
Languages / Data · Python · C/C++ · SQL · PySpark · PostgreSQL · MongoDB · BigQuery
- 🧠 PhD, Machine Learning — University of Surrey (EPSRC scholarship); MS — IIT Kanpur
- 📄 15+ publications · 175+ citations across IEEE, ACM, Springer, Elsevier, SPIE
- 🏛️ Federally-funded R&D — DARPA (MediFor media forensics), NGA (geospatial), DoE (nuclear-safety monitoring), DoD, EU
- 🔍 40+ journal reviews — IEEE TIP, IEEE TSP, CVIU, MDPI, PE&RS
- 📜 US Patent in geospatial intelligence · Co-founder, Ubihere (GNSS technologies)
I'm open to senior AI/ML engineering and applied-science roles (onsite, hybrid, or remote) and interesting collaborations.
📌 Pinned repositories below span my agentic-AI & RAG builds and my computer-vision & medical-imaging research.