Experience

Lead the software engineering department as the sole engineer; own and manage all projects end to end
Designed and built Geocon AI Copilot, a company-wide internal AI assistant (Python/Flask, PostgreSQL, Azure App Service). Owned the full stack: Entra ID SSO, SSE streaming chat over the OpenAI Responses API, RAG over vector stores with intent-based tool routing, Projects workspace with memory/recaps, SharePoint grounding via Azure AI Foundry, admin analytics, and a blue-green GitHub Actions deploy pipeline (~270 tests)
Built Geocon Report Assistant, a privacy-first on-prem RAG system (FastAPI, ChromaDB, Ollama, Entra SSO) over ~1,700 PDFs / ~30,000 pages (~60% scanned). Features resumable OCR ingest, citation-linked PDF viewing with highlighting, cross-encoder reranking, and structured geotechnical metadata for exact filter/count queries; deployed behind Cloudflare Tunnel with RBAC and analytics
Built an AI contract review product for legal: OCR + OpenAI comparison against prior agreements (or baseline risk review), branded PDF export; shipped as both a Flask/Azure web app and a Windows desktop app with local indexing and auto-updates
Automated business processes with Power Automate; built Power BI dashboards for performance metrics

Built a scalable real estate news aggregation system using SerpAPI, BeautifulSoup, and cron jobs to scrape, clean, and store thousands of news headlines daily into a CosmosDB NoSQL database, supporting downstream analysis and trend detection.
Engineered a CLIP-based multimodal classification model to evaluate home images for indicators of property condition (e.g., modern vs outdated interiors) and urbanity (e.g., dense cityscape vs rural background). The model helped automatically tag listings with visual insights for investors.
Scraped, parsed, and de-duplicated over 3,000 NARPM property managers and agents across the U.S., building a unified profile per agent with extracted metrics like portfolio size, years of experience, and regional expertise.
Created a Python-based CLI tool that lets users enter a city and returns the top-ranked realtors based on a weighted score combining number of properties (“doors”) and experience, enabling data-backed recommendations for potential clients.

Engineered a 91% accuracy NLP model to verify any Indian government-issued identity.
Developed a robust anti-forgery application utilizing computer vision technologies and libraries including DocTR, TensorFlow, Keras.
Leveraged ngrok for secure tunneling to ensure safe communication between the API and the script.
Conducted comprehensive testing of the application by designing and implementing an API with Postman, employing advanced techniques such as pixel analysis, metadata verification, colol consistency checks, steganography detection, and noise consistency evaluation.

Automated the analysis of log reports to pinpoint errors within the Measure Analyze Recommend Solve (MARS) tool, a high-performance analytics engine.
Ensured high code sustainability and future scalability by utilizing debugging and version control tools (GIT) extensively.
Used the POSIX Regular Expressions Library along with Apache POI to write 1000+ lines of code which can read 15,000+ log files, producing detailed MS Excel visualizations identifying trend irregularities.
Leveraged agile methodologies to drive the iterative development process.
Contributed to the design and implementation of advanced error detection algorithms, improving the accuracy and efficiency of log analysis by 20% and troubleshooting time by 15%.

Administrated the US bank database to improve reporting capabilities and facilitated data sourcing in US Bank’s Access online portal
Troubleshooting 20+ issues daily; Conducting personalized training sessions for employees on card usage and best practices, leading to a 15% decrease in user errors and enhancing compliance with company policies
Optimized customer support workflows, leading to an 18% increase in response time efficiency, enhancing customer satisfaction and service reliability; Troubleshooted issues and provided support through applications such as SAP Concur, Oracle Financial Cloud, ServiceNow