Nadipalli Jaswanth

• Building an AI-powered internal chatbot with RAG architecture (SharePoint ingestion, vector search) and RBAC for role-scoped knowledge access across teams.
• Developed interactive Power BI dashboards and Power Automate workflows for operational reporting, reducing manual data processing effort significantly.
• Performed large-scale data cleaning and transformation in Excel and improving downstream analytics reliability.

• Built full-stack Student Portal (results, mentoring, internship/project approvals) and T&P platform (one-click applications, branch-eligibility enforcement) serving 2,000+ users with RBAC and JWT-protected routes.
• Designed PostgreSQL schema via Prisma with selective denormalisation, normalised core entities while denormalising high-read tables to avoid expensive joins. .

• Developed scalable Spring Boot backend with optimised database queries reducing latency by 30% and RESTful APIs following enterprise standards.

CorpMind - Enterprise Knowledge Intelligence Platform

Tech Stack

: Python, LangGraph, Qdrant, Neo4j, DeBERTa, FastAPI, Next.js, Docker, Supabase, Groq, HuggingFace

Built a production-deployed multi-agent RAG platform with 11 specialized LangGraph agents coordinating across Gmail, Slack, Drive, and Notion - implementing 6 RAG patterns: Naive, Retrieve-and-Rerank, Graph, Hybrid, Agentic, and Multi-Agent RAG. Integrated Neo4j knowledge graph for expertise mapping and decision trail reconstruction combined with Qdrant vector search for hybrid retrieval; deployed DeBERTa NLI contradiction detection catching document conflicts at 0.99 confidence automatically.

FlowCRM

Tech Stack

: TypeScript, Node.js, Express, Prisma, PostgreSQL, NextJs, Redis, Socket.io

Built multi-tenant CRM with org-level RBAC, JWT + 2FA auth, org-scoped Socket.io real-time chat, Bull async job queues, lead scoring, deal pipeline, and per-org rate-limited AI assistant (Nemotron) with streaming inference. Implemented analytics dashboard with CSV/Excel export, Cloudinary file handling, webhook support, and real-time notifications across isolated tenant environments.

Unified Text-to-Text Transformer (T5-style)

Tech Stack

: PyTorch, Python, NumPy, Transformers

Built custom multitask transformer from scratch with multi-head self-attention, cross-attention, shared encoder and per-task decoder heads; trained on translation, summarisation, and QA with task-conditioned prompts, evaluated via BLEU, ROUGE, and Exact Match