Harsh Dhariwal

Product Leadership: Managed a team of 8 to launch a first-of-its-kind EB-1A legal portal, defining the entire product vision, UI/UX roadmap, and AI feature set from scratch.
Enterprise Architecture: Designed a production-grade Azure CI/CD pipeline with container recovery protocols and automated failure monitoring to ensure 24/7 reliability.
Agentic Automation: Architected a LangGraph-based multi-agent system for 150+ page petition generation, utilizing ChromaDB for intelligent evidence retrieval.
Business ROI: Reduced legal document generation time by 98% (50 hrs → 45 min) and slashed costs by 99.9% ($12,500 → $10)

Product Innovation: Engineered a generative drug discovery platform that compresses 10-year R&D cycles into 2–3 years.
Autonomous Architecture: Architected research agents using Serper.ai and Qdrant for de novo molecule generation.
Technical Validation: Validated chemical viability via QSAR regression modeling and NVIDIA-accelerated molecular docking.

Enterprise Architecture: Engineered a secure, VPC-hosted transcription and audit pipeline utilizing Fine-tuned Whisper and Phi-3.5 MoE.
Scalable Compliance: Developed automated PII redaction and script-adherence protocols, scaling to evaluate 800+ calls daily.
Cost Optimization: Reduced operational expenses to 1 INR per call through model optimization and manual intent-classification tuning

Scalable Architecture: Developed a database-agnostic pipeline achieving 90% fidelity across 15+ domains without code modifications.
Algorithmic Innovation: Leveraged a hybrid stack of CTGANs, Gaussian Copulas, and Semantic Embeddings to preserve complex data correlations.
Data Integrity: Integrated Topological Sorting to ensure 100% referential integrity for high-utility ML training in privacy-restricted environments