Ujjwal
Kashyap

An enterprise-grade, multi-tenant AI assessment platform integrating seamlessly with LMS providers via LTI 1.3. It enables instructors to create complex verbal and roleplay scenarios where learners converse with animated AI personas, backed by real-time speech-to-text and concurrent LLM evaluation pipelines for evidence-based competency reporting. As a major architectural contributor, I led critical performance, security, and infrastructure overhauls that transitioned the product to a scalable enterprise system. This included engineering a custom runtime prompt-management platform, establishing session-credit billing engines, and designing capability-based RBAC systems with safe feature-flagged rollouts.

An AI-powered, multi-tenant fact-verification platform that decomposes large documents into atomic claims and adjudicates them against the live web, knowledge graphs, and large language models. It returns per-claim verdicts with cited evidence, confidence scores, and suggested corrections. As the primary backend architect, I engineered the core cascading verification pipeline to balance LLM cost, latency, and accuracy at scale. I transformed the initial prototype into a high-throughput, concurrent service by eliminating resource leaks, building a comprehensive multi-layer caching subsystem, and implementing strict role-based multi-tenant security.

An AI-driven sales enablement platform that replaces human-led training with scalable, interactive simulations. It utilizes AI buyer bots to emulate real-world sales calls, capturing telemetry to drive automated coaching and analytics. I architected and implemented the core workflow engine for the AI sales-persona training bots. This involved building robust feedback loops capable of concurrently simulating complex negotiation and cold-call scenarios, analyzing candidate performance, and generating actionable coaching insights without human intervention.

A comprehensive ecosystem of applications designed to streamline the content creation and publishing lifecycle. The platform includes a unified CMS, the Dictera Suite, and specialized workflow tools for authors and publishers. I developed and maintained the core backend services powering the Dictera Suite admin panel. My focus was on establishing secure, multi-tenant foundations, implementing sophisticated role and permission matrices, building resilient rate limiters, and ensuring robust session handling for enterprise users.

An AI-driven talent acquisition platform designed to automate and accelerate hiring workflows. It leverages intelligent agents and standardized assessments to drastically reduce manual screening time for recruiters. I led major backend modernization initiatives, transitioning fragmented legacy API stacks into a unified, secure v3 architecture. I built the foundational data ingestion pipelines and integrated specialized AI agents to automate the evaluation of both technical and subjective candidate data.