Hire Manan S. - LLM Engineer

* Building AI capabilities for the TIFIN RM platform, enabling financial advisors to retrieve insights and resolve complex client queries using Generative AI. * Designed and shipped production multimodal RAG systems over financial documents, PDFs, tables, and structured filings * Built agentic AI workflows with multi-step reasoning, tool calling, and grounded response generation * Developed Text-to-SQL systems for querying portfolio and CRM databases using natural language * Finetune, Deployed and optimized open-source LLMs including LLaMA 3, Mistral, and Qwen using vLLM and quantization techniques * Improved response quality through retrieval re-ranking, grounding pipelines, and evaluation frameworks * Built evaluation systems tracking faithfulness, retrieval quality, latency, and hallucination metrics.

* Worked on enterprise conversational AI systems and Generative AI applications powering Orimon, a no-code AI platform used by global clients across the US, EU, and APAC. * Architected and developed core components of Orimon, enabling enterprises to build AI-driven chatbots and virtual assistants * Fine-tuned and productionized open-source LLMs including LLaMA and Mistral for enterprise client deployments * Built conversational AI solutions for clients including Air Arabia, Indigo, Education First, Robi, and Vodafone * Designed and implemented intent detection systems for Oriserve’s NLU engine from research to production deployment * Developed real-time AI-powered email reply generation systems for enterprise customer interactions * Implemented reinforcement-learning-based recommendation systems within sales automation workflows * Built dialogue management systems for multi-turn conversational AI applications * Collaborated across research, product, and engineering teams to productionize AI solutions at scale

* Summer Research Intern under Dr. Ashish Ghosh at the Machine Intelligence Unit, Indian Statistical Institute Kolkata. * Analyzed Self-Organizing Maps (SOM) algorithms with a focus on neighborhood function selection for synthetic datasets * Conducted experiments and comparative evaluations across different SOM neighborhood strategies * Worked on unsupervised learning techniques and neural-network-based clustering approaches