Jun 2025 – Present · 11 mos · Bengaluru, Karnataka, India · On-site

• Designed and deployed multi-agent orchestration frameworks enabling autonomous coordination, context retention, and task-driven reasoning across AI systems.
• Engineered fact-verification and multi-source validation pipelines to reduce LLM hallucinations and enforce verifiable, citation-based output standards.
• Developed memory-aware conversational logic for persistent user context handling, repeat-query detection, and stateful multi-turn dialogue management.
• Built end-to-end agent pipelines for travel and data analysis applications using the Google Places API, integrating sentiment insights and enforcing reasoning consistency.
• Constructed a benchmarking suite for evaluating agent performance across LLMs (GPT, Claude, Gemini, Groq), driving data-informed optimization cycles.
• Enhanced prompt reliability and fallback mechanisms, improving reasoning coherence and system consistency across agent workflows.
• Built a CrewAI-powered multi-agent financial intelligence system composed of domain-specialized expert agents collaborating via shared memory and context graphs to produce unified financial analyses.
• Developed and integrated custom toolchains leveraging Perplexity and Tavily Search APIs for real-time data retrieval, reasoning augmentation, and grounded intelligence generation.
• Implemented an OCR + Vision pipeline combining DeepSeek OCR and Qwen Vision models for image-to-text-to-text understanding, enabling document parsing, visual data extraction, and analytical chart summarization.
• Adopted Model Context Protocol (MCP) to standardize context and tool interoperability across agents, enhancing scalability and modular integration between LLMs and external systems.
• 🧩 Impact: Strengthened AI agent reliability, interpretability, and context-awareness positioning the company’s Agentic AI framework for scalable deployment across enterprise projects.