Mohammad Ali Khan

AI Engineer

Orange Antelopes

Feb 2025 - Current (1 year 5 months)

Designed and deployed LLM-powered agents using LangChain and LangGraph for workflow automation, email intelligence, and decision routing. Built a Construction Scheduling Agent combining LLM reasoning and RAG, incorporating human feedback loops (RLHF-style scoring) to iteratively improve plan quality and task prioritization. Architected multi-vector RAG pipelines over structured SQL data and unstructured documents with metadata-aware retrieval. Containerized and deployed ML services using Docker, and Kubernetes, integrated with Microsoft Azure ML for model management, endpoint deployment, and pipeline orchestration.

E

Data Scientist

Experiential Etc

Jul 2024 - Jan 2025 (7 months)

Led development of YOLOv8-based detection and tracking pipelines (ByteTrack, DeepSORT) for real-time retail analytics across multi-camera RTSP feeds. Designed edge-optimized inference pipelines on Jetson Nano using TensorRT and DeepStream, balancing latency, accuracy, and resource constraints. Implemented event-based counting and demographic analytics using spatio-temporal logic and decision boundaries for reliable production metrics. Built end-to-end ML workflows including data curation, model training, evaluation, and monitoring, tracking experiments with Weights & Biases. Collaborated with product and backend teams to productionize ML outputs into dashboards and APIs with GDPR/DPDC-compliant data handling.

N

Associate Data Scientist

Neural Bridge

Aug 2023 - Jun 2024 (11 months)

Developed and fine-tuned NLP models (BERT, spaCy, Hugging Face) for intent detection, semantic similarity, and classification tasks. Built prototype conversational systems with context memory, evaluation metrics, and deployment-ready pipelines. Taught Machine Learning, Deep Learning, Computer Vision and Reinforcement Learning, emphasizing training workflows, evaluation strategies, and system design thinking.

S

Data Scientist (AI & Computer Vision)

SportAI Tech LLP

Jul 2022 - May 2023 (11 months)

Designed a sequence-based activity recognition system, recognizing that actions unfold over time rather than static frames. Extracted pose trajectories using MediaPipe, converted spatiotemporal features into structured sequences, and trained LSTM-based classifiers for shot recognition. Built learning pipelines combining vision preprocessing, temporal feature extraction, model training, and inference. Integrated trained models into a real-time sports analytics platform, enabling automated event detection and performance analysis.

Mohammad Ali Khan

About

Detail

Timeline

Résumé