Two AI engineers specializing in LLM-powered applications, multi-agent architectures, RAG pipelines, and production-grade ML systems — from chatbots to recommendation engines serving millions.
Architected a next-gen RM portal at Nexedge Capital empowering wealth managers with an AI-powered chatbot for client portfolio analysis, risk assessment, and personalized advisory. Features intent classifiers for query routing, Azure Blob RAG for secure document retrieval across client records, RAPTOR for hierarchical document reasoning, and real-time data converters delivering comprehensive financial insights.
Engineered a sophisticated WhatsApp chatbot at Nexedge Capital using multi-agent architecture for wealth management clients. Specialized agents handle portfolio queries, market research, tax advisory, and document processing orchestrated via LangGraph with intelligent routing. Leverages Gemini and Claude APIs with RAPTOR for retrieval-augmented refinement and Azure Blob RAG converters.
Designed and deployed end-to-end data automation pipelines at Nexedge Capital, transforming manual wealth management workflows into intelligent, event-driven systems. Automated client onboarding data extraction, portfolio rebalancing triggers, regulatory report generation, and cross-system data synchronization using LLM-powered orchestration — reducing operational overhead by 65%.
Built domain-aware conversational AI using Rasa, Ollama, LangGraph, LangSmith, and LangChain-based RAG pipelines with LlamaIndex and Hugging Face. Reduced manual tasks by 70% through LLM-based event triggers, prompt orchestration, and intelligent automation workflows.
Created AI-powered analytics dashboards using Metabase, NocoDB, and Python, delivering NLP-based insights, embeddings-driven recommendations, and predictive analytics. Real-time data pipelines with event-driven APIs powering automated decision systems.
Designed an end-to-end resume intelligence system converting unstructured resumes into structured entities for search ranking, autofill, and recommendations. Fine-tuned BERT models for section segmentation with domain-specific NER pipelines. Achieved 90% improvement in extraction coverage over legacy systems.
Architected a large-scale recommender system processing 6M candidate profiles using dense embeddings, FAISS-based nearest neighbor search, and collaborative filtering. Designed hybrid relevance scoring combining semantic similarity and behavioral signals for personalized job discovery.
Designed an LLM-based resume summarization pipeline using LLaMA-3 processing 3.54 lakh resumes/day at 1000 tokens/sec. Achieved INR 1 crore annual cost savings through batching, prompt optimization, and inference efficiency at scale.
Built a production pipeline combining Whisper (ASR), LLaMA-8B for key phrase extraction, and YOLO-based vision models for automated video framing. Generates 40-50 second highlight clips, improving content discoverability and engagement.
Developed a real-time range prediction model from scratch using noisy telemetry time-series data across 40k electric scooters at OLA. Validated on 2k scooters during Move OS 3 beta rollout, achieving 86% real-world accuracy.
Built a production-grade CV pipeline using DeepLab-V3 for segmentation and VGG16 for classification to detect driver uniform compliance from selfie images at OLA. Achieved 98% precision under real-world noise conditions.
B.Tech. Computer Science Engineering — 7.4 CGPA
Galgotias College of Engineering and Technology
2019 — 2023
B.Tech, Civil Engineering — 7.159 CGPA
Indian Institute of Technology, Roorkee
2018 — 2022
Senior AI Engineer — Gurugram, India
+91 86199 57072
Senior Data Scientist — Bharatpur, India
+91 94610 23175