utkarsh.jain
About Candidate
Software Engineer with around 3 years of experience specializing in Generative AI (GenAI), LLM Orchestration, and
scalable backend systems with expertise in architecting production grade Retrieval Augmented Generation (RAG)
pipelines, Autonomous LLM powered AI Agents, and conversational AI frameworks. Proven track record in the
full AI Product Lifecycle, from system design, prompt engineering, multi-agent orchestration, finetuning to
evaluation, benchmarking, guardrailing and deployment of high concurrency systems. Adept at leveraging Vector
Databases and Semantic Search to build intelligent, data-driven automation. Strategic collaborator focused on
delivering low latency, scalable AI solutions that drive business value and enhance user engagement.
Education
Work & Experience
AI-Powered Recruitment Platform •Designed and implemented the Clarification Agent, an AI assistant for recruiters that captures nuanced job requirements and contextual information through guided dialogues, directly improving candidate sourcing accuracy and talent recommendations. •Developed the Screening Conversation Agent, enabling AI-led candidate interactions via call and WhatsApp to share job details, assess interest, and gather structured screening insights at scale. AI-Based Application Scoring •Contributed to develop a Hybrid Ranking Engine that transitioned the platform from basic Sentence-Transformer model to an Agentic AI Framework integrated with a LightGBM Classifier, significantly enhancing candidate evaluation across multiple dimensions. •Optimized System Performance by achieving a 55% increase in NDCG, 58% boost in MRR, and 31% improvement in CTR, while elevating Precision by 32% and Recall by 45% benchmarked on realtime data.
AI-Powered IVR and Conversational Bot •Architected a Semantic Caching Layer using Vector Embeddings, decreasing system latency by 40% for repetitive LLM queries. •Developed a Model-Agnostic plug and play module, allowing for seamless LLM model switching and A/B testing to optimize for accuracy vs. inference cost. •Worked on AI Assist for Human-in-the-Loop workflows to provide real-time, context-aware assists and domain knowledge to human agents during live calls. •Implemented Agentic Workflows for automated call handling, including Intent Recognition (Auto Reasoning), Entity Extraction (Auto ToDos), and Structured Call Summarization by processing live call streams. AI-Driven Email Assistant •Designed and developed an AI Email Assistant that automates end-to-end processing of incoming organizational emails, extracting queries and generating contextually accurate responses. •Built an evolving RAG knowledge base from policy documents and prior correspondence, enabling consistent and up-to-date answer generation over time. •Implemented a reusable query bank to ensure response consistency across recurring queries, and developed persona-based email response generation tailored to different recipient types.
