• Autonomous Agent System (LangGraph): Engineered a multi-agent framework that orchestrates cross-platform data scraping and real-time anomaly investigation within time-series databases. Developed a Reasoning Agent to autonomously pull event data, execute multi-model anomaly detection, and perform multi-dimensional root-cause analysis (RCA). This system evolved from synthesizing product advice to a mission-critical tool adopted by the PM team for automated daily insights and system health monitoring.

• Interactive Analytics Platform: Developed a high-performance dashboard to visualize social media trends and automated anomaly diagnostic reports. Integrated unsupervised clustering for feedback themes with advanced time-series analysis to detect emerging market signals and database irregularities. Engineered a self-service alerting pipeline that transforms raw event spikes into actionable visual insights, facilitating rapid cross-team response to market and system shifts.

• ML Infrastructure & Optimization: Orchestrated the deployment and fine-tuning of custom-trained models on third-party GPU cloud providers.

• Multi-Modal Agent Architecture: Orchestrated a suite of perception tools within a LangGraph framework, integrating YOLO for object/person tracking, Whisper for speech transcription and VLM analysis, effectively empowering LLMs to "see" and "hear" raw video data.

• Dynamic Query Routing: Designed an intelligent decision-making layer that classifies user intent and dynamically routes queries to specialized sub-agents. This optimized system performance by handling simple tasks locally while reserving complex reasoning for advanced tool chains.

• Semantic Video Retrieval: Built a searchable media library enabling users to query video content via natural language. Implemented vector-based indexing to allow precise retrieval of specific video segments and timestamps based on visual and audio keywords.

• Multi-Modal Agent Architecture: Orchestrated a suite of perception tools within a LangGraph framework, integrating YOLO for object/person tracking, Whisper for speech transcription and VLM analysis, effectively empowering LLMs to "see" and "hear" raw video data.

• Dynamic Query Routing: Designed an intelligent decision-making layer that classifies user intent and dynamically routes queries to specialized sub-agents. This optimized system performance by handling simple tasks locally while reserving complex reasoning for advanced tool chains.

• Semantic Video Retrieval: Built a searchable media library enabling users to query video content via natural language. Implemented vector-based indexing to allow precise retrieval of specific video segments and timestamps based on visual and audio keywords.

• Managed multi-user HPC platform supporting 20+ concurrent AI/ML workloads

• Designed resource monitoring and job scheduling system for optimal cluster utilization

• Implemented containerized environments using Docker Swarm for reproducible research

• Built backend APIs for HPC management platform improving operational efficiency

• Administered Linux servers and storage infrastructure for research teams

Tech Stack: Linux, Docker, Python, HPC, Distributed Systems, Infrastructure Management

• Architected scalable data ingestion pipeline processing 100GB+ agricultural data daily with integrated OCR services

• Designed and implemented RESTful APIs handling 10K+ requests/day using Django REST Framework

• Built microservices architecture for ML model serving, reducing inference latency by 40%

• Developed distributed data processing system for knowledge graph with 100M nodes

• Managed deployment infrastructure using Docker and Kubernetes

Tech Stack: Python, Django, PostgreSQL, Docker, Kubernetes, Microservices, REST APIs