GOVINDA REDDY/ AI ENGINEER
● BUILDING PRODUCTION AI SYSTEMS HYDERABAD · IN EST. 2025

Kurri
Govinda Reddy

AI / ML Engineer / Software Engineer (AI)

I build production AI agents, LLM systems, and multi-agent pipelines — from an enterprise underwriting & fraud engine to an autonomous code-generation system and an open-source legal-operations agent.

View Work → Resume ↓ GitHub LinkedIn
govindareddy@dossier ~ %
$ whoami --role
01

Summary

AI / ML Engineer specializing in production AI agents, Large Language Models, and multi-agent systems. I've shipped an enterprise insurance underwriting and fraud-detection engine, an autonomous multi-agent code-generation system, a multi-tenant RAG chatbot, and an open-source AI legal-operations agent.

Hands-on with custom LLM routing across 10+ providers (Mistral 7B, LLaMA 3.2, Claude 3.5 Vision, Qwen 2.5 Coder, Gemini, Groq, Ollama), agent tooling, function calling, RAG over pgvector, document-processing pipelines, and end-to-end FastAPI / SQLAlchemy services.

Currently going deep on enterprise AI security & governance — AI gateway architecture, identity (OIDC / SSO), audit logging, and incident response — grounded in building Vouchsafe, an open-source AI compliance scanner mapped to the EU AI Act, ISO/IEC 42001, and NIST AI RMF.

Focus
AI Agents · LLMs · RAG · AI Security
Education
B.Tech CSE (AI & ML)
Email
govindreddy.kurri@gmail.com
Phone
+91 94942 18866
Location
Hyderabad, IN
0%
Code-gen quality · manual review
0×
Faster F1 · Groq vs Gemini
0+
LLM providers routed
0%
Precision · CUAD key clauses
02

Stack

AI / ML / LLM

LLMsAI AgentsMulti-Agent SystemsLangChainLangGraphLangSmithRAGpgvectorEmbeddingsVision Language ModelsPrompt EngineeringFunction CallingAgent ToolingHugging FaceMCPGenerative AI

Backend & APIs

FastAPIPydanticSQLAlchemyAlembicREST DesignAsyncWebSocket / SSEOpenTelemetry

AI Security & Governance

AI GovernanceEU AI ActISO/IEC 42001NIST AI RMFOWASP LLM Top 10MITRE ATLASAI Gateway PatternsLLM Routing & FailoverOIDC / SSO FundamentalsAI Incident ResponseAudit LoggingLLM Cost Observability

Languages & Data

PythonSQLPostgreSQLMySQLMongoDB

Cloud / DevOps / Frontend

GCPAWSAzureDockerCI/CDGitHub ActionsTypeScriptAngular MaterialLinuxBashruffmypy
03

Experience

Monocept Jan 2026 — Present

Software Engineer

AgentFX — Agentic Claims & Fraud Automation (Insurance)
STACK · Python 3.12 · FastAPI · Pydantic v2 · LiteLLM · PostgreSQL + pgvector · SQLAlchemy · Alembic · scikit-learn · Angular 17
  • Architected a config-driven agentic framework where each customer is onboarded and tuned from configuration alone — new tenants, routing, rule packs and pipelines go live without redeploying.
  • Built the claims fraud-detection pipeline as cooperating agents that hand off agent-to-agent — extraction → aggregation → multi-dimension scoring — to reach a single final fraud verdict with rationale.
  • Added document forensics — tampering detection (EXIF / ELA / PDF-structure) and AI-generated document detection — to catch manipulated or synthetic claim documents before straight-through processing.
  • Built the RAG stack: watcher-driven ingestion, pluggable chunkers, hybrid dense (pgvector) + BM25 retrieval fused via RRF, cross-encoder rerank, and grounded citations.
  • Built a call-intelligence pipeline — transcription → analysis → quality and lead scoring — as a pluggable agent pack.
  • Hardened LLM access behind a provider-agnostic AI gateway: only authorized consumers can reach the models — any other machine that pings the LLMs is denied (per-consumer keyed access).
NeuralShell — Autonomous Multi-Agent Code Generation
STACK · Python 3.11+ · FastAPI · WebSocket · Ollama (qwen2.5-coder) · MySQL · Docker · Textual (TUI) · ruff · mypy
  • Built NeuralShell, an autonomous multi-agent system for code writing, review, testing, and deployment — code-gen quality rated ~90% via manual review.
  • Introduced an ArchitectureNode for intelligent codebase parsing and change-location detection before any modification.
  • Implemented parallel agent nodes (Code / Review / Testing / Deploy) coordinated by a WorkflowRouter that dynamically delegates tasks.
  • Designed custom agent tools (file read/write, codebase navigation, context retrieval) exposed via function calling, with a self-updating context file that preserves project awareness across runs.
  • Shipped a Textual TUI client with interactive mode, slash commands, dry-run previews, and target scoping — deployable across any codebase.
Monocept May 2025 — Jan 2026

Software Intern

MozartBot — Multi-Tenant RAG Chatbot
STACK · Python · FastAPI · LangChain · LangGraph · PostgreSQL · pgvector · TypeScript · Angular Material · Ollama
  • Architected a multi-node LangGraph workflow (retrieval → grading → generation) with semantic + late chunking and checkpoint-based memory over pgvector — document Q&A in 5–7s.
  • Built a multi-format document pipeline (PDF, DOCX, Excel, CSV, image OCR) with PyMuPDF multi-page table extraction and token-aware chunking into a multi-tenant pgvector store with per-company database isolation.
  • Integrated 8–10 LLM providers (primarily Mistral 7B & LLaMA 3.2) with hot-swappable selection — ~80% retrieval accuracy across 8 conversation flows.
  • Implemented SSE streaming across a multi-endpoint REST surface with async connection pooling and Alembic migrations, plus a responsive Angular Material frontend with live chat and upload progress.
04

Selected Projects

ScriptOCR

Open Source

A multi-script OCR engine that keeps its coordinates. Auto-detects the script in each region and routes it to a dedicated recogniser — 6 languages (English, Hindi, Telugu, Kannada, Tamil, Malayalam), printed and handwritten — then extracts key/value pairs and tables from page geometry alone (no LLM). Trained the handwriting and Indic recognisers in-house; packaged on ONNX Runtime for CPU-only, fully offline / air-gapped deployment.

ONNX Runtime6 languageshandwritingKV + tablesCPU-onlyoffline
↗ Repo

Vouchsafe

Live · v0.3.0

A signed, deterministic inventory of every AI / LLM / agent / MCP touchpoint in a codebase — mapped to EU AI Act, ISO 42001, NIST AI RMF, OWASP LLM Top 10 & MITRE ATLAS. No LLM inside. Air-gapped. Byte-identical output, signed with Ed25519.

PythonEd2551915 languages45+ AI SDKsGitHub Action
↗ Repo↗ Marketplace

Legal Co-Pilot

Open Source

A multi-provider AI legal-operations agent on a $0/month budget. Ingests contract PDFs and runs the full lifecycle: extraction → risk analysis → redlining → drafted negotiation email → action layer → long-term memory. Benchmarked on the public CUAD dataset.

FastAPIpgvectorPydantic v2GeminiGroqOllamaMCP
↗ Repo
05

Education & Certifications

Education
Vasireddy Venkatadri Institute of Technology
B.Tech in Computer Science & Engineering — Artificial Intelligence & Machine Learning
2021 — 2025
Narayana Junior College
Intermediate (MPC) — Mathematics, Physics & Chemistry
2019 — 2021
Loyola Public School
ICSE — Class X
2018 — 2019
Certifications
  • Associate Cloud EngineerGoogle Cloud
  • Generative AI Essentials ProfessionalMicrosoft
  • Data Analysis Essentials ProfessionalMicrosoft
  • AI for EveryoneDeepLearning.AI
  • Docker Foundations ProfessionalDocker
06 — Let's build

Let's make something
that ships.

Open to AI / ML engineering roles and collaborations on agents, LLM systems, and applied ML.

govindreddy.kurri@gmail.com +91 94942 18866
GitHub LinkedIn Resume ↓ Email →