Skip to content
View Bruhadev45's full-sized avatar
🎯
Focusing
🎯
Focusing
16 followers · 16 following

Achievements

Achievement: YOLOAchievement: Pull SharkAchievement: Pair ExtraordinaireAchievement: Quickdraw

Achievements

Achievement: YOLOAchievement: Pull SharkAchievement: Pair ExtraordinaireAchievement: Quickdraw

Highlights

  • Pro

Block or report Bruhadev45

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Bruhadev45/README.md

Kandimalla Bruhadev

AI Engineer — LLM Systems, RAG & Agents · Published IEEE Researcher · Oracle Certified ×2

Portfolio LinkedIn Email

About

I build grounded AI systems — retrieval pipelines, LLM agents, and the production software around them.

  • AI Automation Engineer at InvisiEdge Marketing — AI-powered automation tools and platforms for enterprise workflows
  • Published researcher — first author of an IEEE paper on JurisGPT, a citation-grounded legal assistant for Indian startup and corporate law
  • 100+ LLM evaluation tasks at Outlier AI — RLHF workflows, prompt optimization, and code-generation review
  • Client work shipped to production — most recently Vectius, a SaaS operations platform with an encrypted credential vault, invoice OCR + LLM autofill, and a natural-language assistant

Research

JurisGPT: A Citation-Grounded Legal Assistant for Indian Startup and Corporate Law — IEEE Live at jurisgpt.me · Code

RAG over a maintained corpus of 47,867 Indian legal documents (Companies Act 2013 + nine further statutes). Hybrid BM25 + lexical retrieval fused with weighted Reciprocal Rank Fusion; every answer carries inline citations, calibrated confidence badges, and limitation statements.

  • Hybrid retrieval reaches Recall@5 0.84 / MRR 0.95 on a 120-query human-verified benchmark (Fleiss' κ = 0.81)
  • 80% of answers judged citation-faithful, 2.5% severe-hallucination rate
  • Negative result with practical value: a generic MS MARCO cross-encoder reranker lowers nDCG@5 by 8.7 points on hybrid (p = 0.0002) and 16.3 on dense retrieval (p = 0.0001) — paired permutation tests over the full benchmark, reproducible byte-for-byte in rag-eval-lab. Retrieval quality in legal RAG is corpus-bound, not retriever-bound

Featured Projects

JurisGPT · live

Citation-grounded legal RAG assistant for Indian startup law. Published as an IEEE paper — see Research above.

RAG FastAPI Next.js Legal AI

Vectius · client work

SaaS operations platform: AES-256-GCM credential vault, invoice OCR + LLM autofill with approval chains, spend analytics, GPT-4o-mini assistant.

SaaS Security LLM Extraction

AI-Powered HRMS · live

HR platform with AI resume parsing, candidate–job matching, and a RAG-based HR chatbot. Next.js 15 + Turso.

Next.js 15 RAG Turso

rag-eval-lab · benchmark

Reproducible retrieval benchmark behind the reranker finding: paired statistics, seeded bootstrap CIs, and CI that fails if make reproduce drifts a single byte.

Evaluation Statistics Reproducible Research

ragfuse · library

Weighted Reciprocal Rank Fusion for hybrid retrieval — real per-retriever weights, provenance on every hit, zero dependencies. 100% test coverage, mypy --strict.

Python Information Retrieval RRF

ToolLedger

Multi-tenant SaaS for credential management and subscription analytics. Encrypted vault, tenant isolation at the data layer, NestJS + Prisma + PostgreSQL.

SaaS Multi-Tenant NestJS

Tech Stack

AI / ML Python LangChain PyTorch Hugging Face OpenAI Anthropic

Full Stack TypeScript Next.js React FastAPI Node.js

Data & Infra PostgreSQL Chroma Docker AWS Vercel

Certifications

Oracle Oracle

GitHub Stats

Open to AI engineering roles and consulting — LLM systems, RAG, and agents

Portfolio

Pinned Loading

  1. ai-powered-hrms ai-powered-hrms Public

    AI-powered HRMS with resume parsing, candidate-job matching, and a RAG-based HR assistant. Next.js 15, Drizzle, Turso.

    TypeScript 1 1

  2. Juris-GPT Juris-GPT Public

    Citation-grounded legal RAG assistant for Indian startup and corporate law. Hybrid BM25 + weighted RRF retrieval over 47,867 statutory documents. IEEE-published.

    TypeScript 3 1

  3. FinSight-X-AI FinSight-X-AI Public

    AI financial-insights assistant. Extracts metrics, sentiment, and risk from filings and news with RAG grounding. Next.js + GPT-4.

    TypeScript

  4. openenv-incident-response openenv-incident-response Public

    SRE incident-response training environment for AI agents. OpenEnv-compatible, three progressive scenarios with quantitative scoring.

    Python

  5. Tool_Ledger Tool_Ledger Public

    Multi-tenant SaaS for secure credential management and subscription analytics. Next.js + NestJS + Prisma + PostgreSQL with an encrypted vault.

    TypeScript

  6. ragfuse ragfuse Public

    Weighted Reciprocal Rank Fusion for hybrid retrieval. Real per-retriever weights, provenance on every hit, zero dependencies.

    Python