Kolli Krupa Johnson Chowdary

• BeverageOS (Tech Lead): Designed the PostgreSQL schema for a beverage POS and kitchen display platform, covering modifier groups, ingredient ownership, and kitchen-display JSON; modeled inventory and recipes with 245 real supplier items and led the move to a NestJS modular monolith.
• WorkStream Platform: Delivered 1 internal platform to a sister company that replaced 3 disconnected tools, covering task management, team hierarchy, document versioning, and messaging.
• LLM Observability Stack: Improved LLM inference reliability by instrumenting 3 production metrics with structured logging and live health dashboards for monitoring and troubleshooting.
• Pixel Streaming Platform: Cut project onboarding from hours to minutes by engineering a 4 stage provisioning pipeline with 3 tier RBAC and session queue management.
• LMS Teacher Dashboard: Removed administrative bottlenecks across 3 educator workflows by centralizing class management, assignment tracking, and quiz assessment.
• CAD Comparison Tool: Eliminated manual drawing review across 3 change types by building a 4 stage pipeline covering parsing, diffing, visualization, and PDF reporting.
• Distro Streaming Ecosystem: Enabled secure kiosk streaming across a 2 component architecture using a Flutter Android client and virtualized server backend.

Tools: Python, AutoGen, LangChain, Ollama, ChromaDB, Redis, MCP, tree-sitter, FastAPI, Docker, Git, PyPI

• Multi agent orchestration: Coordinated a 6 agent AutoGen pipeline over a 17 tool registry, with 3 tier permissions controlling which tools each agent can call, and an auto-repair loop with up to 3 fix attempts.
• RAG based agent memory: Removed context loss between agent runs by building a retrieval layer over ChromaDB embeddings with Redis session caching.
• Testing and integration: Validated platform health with a 35 check diagnostic suite and exposed agent tools to external clients through an MCP server.
• ZonnyHelper CLI: Reduced deployment to a single command across 12 targets by publishing a PyPI CLI with tree-sitter code parsing, Git automation, and fallback across 4 LLM providers.

Tools: Python, LangGraph, Groq API

• Conversational agent workflow: Built a complaint management system as a LangGraph state graph backed by Groq hosted LLMs for an AI engineering take home assignment.

Tools: Python, Memgraph, TimescaleDB, PostGIS, Qdrant, Redis, MinIO, Docker

• LLM assisted classification: Attained full audit traceability across 4 decision types via a 3 tier pipeline of deterministic rules, fine tuned ML, and async LLM escalation.
• Multi store data architecture: Reduced query latency on high throughput workloads by designing a 5 store layer, including Qdrant vector search, each store selected for its access pattern.

Tools: Python, Streamlit, Plotly, spaCy, Transformers

• Natural language analytics: Turned plain-English questions about an uploaded CSV into interactive charts, with data profiling and quality checks built in.

Tools: Python, Flask, Flutter

• End to end diagnosis app: Enabled crop disease and pest identification for farmers through a Flutter photo capture app paired with a Flask admin review dashboard.

Tools: Python, PyTorch, YOLOv8, FedAvg

• Privacy preserving model training: Sustained convergence across non IID distributions by fine tuning YOLOv8 on distributed nodes using FedAvg with gradient clipping and noise injection.

Tools: TensorFlow Lite, MobileNet SSD, OpenCV, Raspberry Pi

• On-device detection: Ran real-time object detection on a Raspberry Pi 4 with no cloud, driving motors directly from what the camera sees.

Tools: Linux, Docker, Nginx, PostgreSQL, MinIO, Portainer, SSH

• Self operated production infrastructure: Preserved continuous uptime across 4 containerized services by configuring Nginx reverse proxy on a fully SSH managed Linux server.

Tools: Arduino (C++), Python

• Embedded control: Built an Arduino robot with 3 IR sensors that remembers its last turn to cross gaps and spins in place on sharp corners.

• CGPA: 9.1 out of 10.0 · Graduated 2026
• Coursework: Machine Learning, Deep Learning, Computer Vision, Cloud Computing, Distributed Systems, Data Structures and Algorithms, Database Management, Operating Systems.