Arshveer – AI agent development, Python, LLM

An AI-powered lead discovery platform that helps financial advisors reclaim over 100 hours each month by automating prospect identification and surfacing data-driven insights to prioritize and convert high-intent prospects faster.

• Architected data pipelines processing 18M+ professional records using AWS RDS, EC2, and PostgreSQL, implementing lead scoring algorithms to surface high-intent prospects for financial services clients.
• Built autonomous agents for web scraping and OSINT data collection using Python, Beautiful Soup, and LLM-augmented workflows to enrich prospect profiles with real-time signals.
• Led DNCL compliance implementation and SOC2 readiness initiatives to ensure enterprise-grade data governance for regulated industries.

Tailored automation solutions for clients in the accounting, legal, and real estate sectors that free up to 30 hours per week by streamlining document verification, sales outreach, client engagement, social media marketing, and customer service workflows.

• Designed and implemented AI-driven automation solutions for clients in the Accounting, Legal, and Real Estate industries.
• Scoped client workflows to identify high-impact automation opportunities across document verification, sales outreach, client engagement, social media marketing, and customer service.
• Architected end-to-end solutions that reduced manual workloads by up to 30 hours per week.
• Managed client relationships from discovery through deployment, translating business requirements into technical specifications and delivering bespoke systems tailored to each firm's operational needs.

This product helps nonprofits engage with supporters, track donations, and manage volunteers—all from a single, easy-to-use platform. Built for Members of Parliament.

FileAssure helps streamline every aspect of your practice, from client intake to document management, with all-in-one software for Licensed Insolvency Trustees.

• Architected enterprise-scale document object detection and PDF text extraction pipelines using YOLO, Tesseract, and Flask, processing financial documents with 97% accuracy.
• Engineered robust ETL pipelines using Airflow to orchestrate high-volume document processing (100K+ documents), to create document-to-insights workflows for financial compliance.
• Integrated LlaMa 3 into legacy systems to enable context-based case routing and file organization.
• Built on-prem infrastructure capable of serving machine learning and language models while staying compliant with SOC2 and PIPEDA.

Computer vision and NLP models that powered a marketing compliance platform across all 50 US states and Instagram; processing thousands of marketing materials (images, videos, audio, and text) and eliminating manual review bottlenecks for cannabis industry clients, reducing review compliance review times from 2 months to 2 minutes.

• Developed computer vision models (ResNet, YOLO) for automated detection of cannabis products and paraphernalia, achieving 92% accuracy, optimizing for recall to minimize compliance risk.
• Improved model performance by 20% through systematic analysis of data collection and annotation pipelines, establishing standardized frameworks that reduced labeling inconsistencies and improved training data quality.
• Partnered with legal and product teams to translate model performance metrics into actionable risk assessments, delivering monthly stakeholder reports that strengthened organizational compliance posture.