Zeeshan Abbasi

At Python Technologies, I build multi-tenant AI products including HR Toolkit AI, Cardinal EA, and Sara AI, taking systems from data models and FastAPI services through LLM workflows and Next.js frontends.

I design RAG pipelines, agentic systems, and integrations that are safe in production: fail-closed grounding, tenant isolation, approval gates, encrypted tokens, HMAC-verified webhooks, and deterministic fallbacks. My work includes Stripe-gated document generation, Slack-delivered executive assistants, and creator assistants integrated with Instagram, WhatsApp, Google Workspace, Microsoft Graph, Zoom, and scheduling tools.

I work across Python, FastAPI, PostgreSQL, pgvector, LangChain, LiteLLM, React, and Next.js, with every AI call routed for cost attribution and testable offline. I focus on enforcing safety and correctness in code rather than prompts.

I also build deep learning systems: SONAR SHIELD delivered 8–15 dB noise-specific SNR improvements and a satellite-error decoder that outperformed Viterbi FEC by 30–40% under burst noise. My projects include autonomous data science, legal-document RAG, and meeting intelligence with Whisper.