vidit manglani

About

Built and productionized an LLM-enabled system visualization web app at Longshot, designed modularly to pivot quickly to a mobile-friendly interface as customer needs changed. Experienced in diagnosing LLM/agentic workflow failures using observability, deterministic replay, and fault-tree root cause analysis. Also delivers developer-focused demos and trainings (including robotics deployment/mapping at Meta) and partners with sales as a technical closer, including for government clients by demonstrating failure modes and system modularity.

Key Strengths

• Prototype-to-production execution with iterative delivery, testing, and documentation
• Customer collaboration to clarify requirements and confirm understanding in writing
• Modular architecture enabling fast pivots (e.g., adding mobile-friendly UI)
• Real-time debugging approach: triage (infra/tools/data), logs/observability, reproduction, deterministic reruns, fault-tree isolation
• Uses deterministic replay to resolve LLM-generated inconsistencies by comparing outputs to expected behavior
• Strong technical demo/workshop facilitation for developer audiences (architecture, data flow, failure modes, reproducibility)
• Enables team autonomy by teaching systems through interactive demos including failure cases
• Acts as technical closer with sales: scenario-driven demos, handles technical objections, builds trust (incl. government clients)

Experience

Education

Skills

Languages