Experience
Associate Consultant (GenerativeAI/Data Science/Software Development)
Kiya.ai
I have engineered Generative AI platforms that reduced deployment time by 60%, scaled to 2,500+ users, and boosted automation by 50% through RAG chatbots and LLM workflows. I also designed FinTech AI pipelines improving anomaly detection accuracy by 35% and developed a multilingual UPI app supporting 12+ languages for secure, scalable digital banking.
- Engineered a no‑code Generative AI Admin Portal enabling clients to build and deploy RAG‑based chatbots with intelligent workflows in minutes. Designed modular microservices for multi‑format ingestion (PDF, DOCX, CSV, URLs), automated data validation, and context‑aware training with multi‑output formats (text, images, source traceability). Integrated open‑source (Mistral, LLaMA) and commercial LLMs (OpenAI, Anthropic), with CrewAI and LangGraph agents for dynamic query classification and parameter extraction. Deployed on GPU‑accelerated AWS infrastructure, achieving 2,500+ concurrent users with low‑latency response times and 50% higher automation coverage.
- Engineered an AI log analyzer converting banking system logs into natural language summaries. Developed fine‑tuned LLM pipelines for anomaly detection, boosting accuracy by 35%. Delivered as secure FastAPI microservices, ensuring compliance in FinTech environments.
- Developed a multilingual voice‑enabled UPI app live across Indian public sector banks. Supported 12+ languages, secure payment workflows, and context‑aware responses. Leveraged SBERT, NER, and zero‑shot classification with REST APIs on Azure Cloud.
Data Scientist
Kiya.ai
As a Data Scientist, I transformed chatbots into RAG‑powered Generative AI assistants, improving response accuracy by 25%, workflow efficiency by 40%, and reducing latency with LangChain, LlamaIndex, and vector databases. I also engineered NLP pipelines and integrated AI solutions across Web2, Web3, and Metaverse, collaborating with 25+ engineers to cut release cycles by 30%.
- Upgraded rule‑based chatbots into contextual SBERT‑based QA systems, boosting response accuracy by 25% and workflow efficiency by 40%.
- Introduced RAG pipelines with LangChain, LlamaIndex, FAISS, and ChromaDB, significantly enhancing contextual relevance and reducing response latency in production.
- Developed conversation summarizers for sentiment and intent tracking, increasing qualified lead capture and enabling data‑driven business targeting.
- Integrated AI assistants into Web2, Web3, and Metaverse ecosystems (Unity, Unreal Engine) via FastAPI, delivering immersive, real‑time interactions across platforms.
- Drove collaboration with 25+ cross‑functional engineers, streamlining deployment workflows and cutting release cycles by 30%.
Research Analyst Intern
Findem
Improved candidate classification accuracy by 15% through AI‑driven automation, engineered Python workflows boosting preprocessing by 40%, and drove 15% higher lead conversions with targeted outreach.
- Analyzed 10,000+ candidate–organization records to refine a classification model, improving accuracy by 15% and enabling more effective candidate-job fits on Findem's 3D Data platform.
- Supported deployment of an AI-driven classification bot, accelerating model rollout and reducing manual review time.
- Built an optimized Python-based data extraction workflow, boosting preprocessing efficiency by 40%.
- Segmented prospect database using advanced analytics to enable targeted outreach, resulting in 20% more focused engagement and a 15% uplift in lead conversions.