I build AI systems that solve real operational problems, from enterprise integrations and document workflows to agent evaluation and automation. My work spans hands-on engineering, working directly with customers, and taking ideas from an ambiguous problem to a deployed product.
I'm a Forward Deployed Engineer at Procore, working within Datagrid, and a USC Computer Science master's graduate. Previously, I've worked across machine learning research, cloud automation, healthcare data, and enterprise software.
- Production AI agents: orchestration, tool use, retrieval, evaluation, and reliable execution.
- Enterprise integrations: MCP tooling, APIs, authentication, and document and construction data systems.
- Applied machine learning: computer vision, multimodal applications, and graph neural networks.
- Full-stack products: Python backends, React/TypeScript interfaces, and cloud infrastructure.
I build and deploy AI agents and integrations for enterprise construction workflows, working directly with customers from discovery through implementation.
- Piping estimation: built man-hours and step-out MCP tooling that reduced a manual estimation workflow from 200–2,000+ hours per job to under 10 minutes.
- Enterprise document integration: delivered a working Documentum connector, resolving undocumented API behavior and non-standard authentication to enable document workflows and agents.
- Construction data: shipped an Autodesk Construction Cloud relationships feature that lets users link assets, sheets, and NavVis files themselves.
- Applied computer vision: annotated 100+ images, trained a Roboflow detection model, and integrated it into a working first iteration of a symbol-recognition and fixture-takeoff workflow.
- Rapid prototyping: built a first working safety-conversation agent in about an hour after discovery, then tested and demonstrated it at an executive meeting. Also built procurement, material-log inspection, and part-code workflows.
- Agent quality and reuse: built concurrent test orchestration with retries and structured outputs; generated evaluation question sets for 15 marketplace agents; created tooling to transfer agents between teamspaces.
- Customer and team enablement: worked onsite, led training sessions, and debugged API integrations with customers and colleagues to turn workflow ideas into working systems.
December 2024 – October 2025
- Built interfaces for wearable technology research and sensor-data analysis.
- Contributed to WaveGNN, modeling irregularly sampled multivariate time series with temporal encoders and dynamic sensor graphs.
- Benchmarked against LSTM, GRU, and GCN baselines; optimized PyTorch training pipelines and hyperparameter tuning.
- Built Python automation and ETL pipelines on Azure, reducing CPU utilization by up to 79% and operational costs by 60–70%.
- Developed VM usage forecasting, scheduling, and monitoring workflows to improve cloud resource efficiency.
May 2023 – October 2024
- Built Tableau and Power BI dashboards and analyzed operational data to improve financial aid workflows.
July 2022 – December 2022
- Developed enterprise applications using React, Java, and Spring Boot.
- Built NLP and OCR workflows for document extraction and automated text processing.
January 2020 – July 2022
- Built Python ETL pipelines, clinical text-processing workflows, and reporting automation for healthcare data.
- Integrated applications with PostgreSQL and MongoDB and optimized data retrieval.
- Sim Studio: integrated image generation, Amazon S3, and Mem0; built usage analytics and supported customer integrations.
- Puter: developed context-aware AI service redirection and asynchronous stream processing in Node.js.
- Headstarter: built and deployed AI products through an engineering residency and buildathons.
| Project | What I built | Stack |
|---|---|---|
| Clipify-AI | Multimodal video analysis with transcription, chaptering, Q&A, search, and keyframe extraction | GPT-4, Whisper, vision models |
| ConversAI | Real-time conversational AI across text, audio, and video, with sentiment analysis and multi-agent support | React, TypeScript, Node.js, OpenAI, LiveKit |
| Compose AI | Chrome extension for real-time AI text completion with a customizable dashboard | JavaScript, Next.js, LLMs |
| CodeConverse AI Editor | AI-assisted code analysis, debugging, and explanations inside an online editor | AI developer tooling |
| Traffic Forecasting with DCRNN | Spatiotemporal traffic forecasting using a diffusion convolutional recurrent neural network | Graph neural networks, time series |
Languages & web: Python · TypeScript · JavaScript · SQL · React · Next.js · Node.js · FastAPI · Django · Flask
AI & ML: LLMs · MCP · RAG · LangChain · LlamaIndex · PyTorch · TensorFlow · Hugging Face · YOLO · Roboflow
Data & infrastructure: PostgreSQL · MongoDB · Redis · Pandas · AWS · Azure · Docker · Kubernetes · Terraform · GitHub Actions
- University of Southern California — M.S. in Computer Science, 2024 · GPA 3.84/4.00
- MIT World Peace University — B.Tech. in Computer Science and Engineering, 2022 · GPA 9.46/10.00
- Impact of Image Translation using Generative Adversarial Networks for Smoke Detection — ComPE, 2022. Combined image translation and object detection for nighttime wildfire smoke detection.
- Evaluation of Fusion Techniques for Multi-modal Sentiment Analysis — ASSIC, 2022.
Manchester United supporter, startup enthusiast, and open-source contributor. Always interested in talking about useful AI products, developer tools, and turning complex workflows into software people enjoy using.