Akash Nallagonda.
AI Engineer.
I design and ship AI systems that connect language models to real data, real workflows, and real users, across Azure OpenAI, Azure AI Search, Model Context Protocol, FastAPI, Next.js, and cloud infrastructure.
Headline metrics
46,000+
students and faculty served
Campus AI assistant Akash builds on at University of Cincinnati.
~$80,000
projected annual cost avoidance
TAMALE extraction gateway, running on a ~$13/month Azure footprint.
30%
lower response latency
Through retrieval and prompt router tuning on BearcatGPT.
20+
concurrent AI initiatives tracked
By leadership through the meeting intelligence platform.
200+
faculty and staff trained
On GenAI and Microsoft Copilot.
150+
students in responsible AI workshops
High school cohort at the RAISE AI Summit.
Best Security Hack
MakeUC 2024
Secure Sight, real time weapon detection platform.
IEEE ICCCI 2024
Published author
Real time student tracking and responsive design papers.
Featured AI systems
Production work, not demos
Four shipped systems chosen to show range: enterprise RAG, document AI middleware, an automation pipeline that replaces a vector database, and a full stack AI product.
BearcatGPT
Campus scale AI assistant and agent platform
Production generative AI assistant serving 46,000 plus students and faculty at University of Cincinnati. RAG pipelines on Azure AI Foundry with hybrid retrieval, Socratic tutoring agents, agentic routing, and MCP integrations to external enterprise systems.
46,000+
Users
30%
Latency reduction
TAMALE AI Extraction Gateway
Enterprise RMS integration on a $13 per month footprint
Stateless FastAPI middleware that bridges Advent TAMALE and BearcatGPT. In memory bitstream extraction turns binary research attachments (PDF, DOCX, XLSX, PPTX) into structured JSON envelopes for LLM tool calls, hardened with API key auth, Key Vault secrets, SHA256 hashing, and standards compliant error codes.
~$80,000
Projected annual cost avoidance
~$13/mo
Azure footprint
AI Meeting Intelligence Platform
Executive use case tracker over Microsoft Teams transcripts
Five layer enterprise workflow that ingests Teams .vtt transcripts, generates strictly grounded JSON via Azure AI Builder, persists to a flat SharePoint list, and exposes the corpus to BearcatGPT through Microsoft Graph. Deliberately skipped a vector database in favor of a per person and date range access pattern.
20+
AI initiatives tracked
$0
Vector DB cost
LectureMind
Full stack generative AI study platform
Generative AI learning platform that turns long form lectures into grounded outlines, summaries, flashcards, quizzes, mind maps, and faculty accessibility reports. Multi model orchestration with fast and reasoning paths, separate student and faculty workspaces, and a transient faculty session mode with cleanup on signout.
Cloudforce No Resume Required
Hackathon
Fast + reasoning
Models routed
What I build
Four shapes of AI work I ship
Different stacks, same engineering bar. Each of these maps to a project on this site.
RAG and retrieval systems
Hybrid search over enterprise corpora with chunking, embeddings, BM25, and re ranking. Grounded answers with citations and refusal guardrails.
LLM agents and tool use
Multi agent routing, controlled handoffs, tool calling with structured outputs, and Model Context Protocol servers connecting agents to real systems.
Automation and workflow pipelines
Production automations that move work between SaaS, Microsoft 365, and LLMs on a schedule or trigger, with logging and human readable outputs.
Full stack AI products and applied ML
Next.js front ends, FastAPI workers, Postgres, and Azure infrastructure delivered end to end, with classical ML and computer vision where they fit.
Proof of work
Published, presented, recognized
Conference papers, leadership talks, hackathon wins, and education work that backs up the engineering.
- Publication2024IEEE ICCCI 2024
An AI Based Student Tracking System to Analyze Student Behavior
Real time classroom behavior analysis and automated attendance using YOLOv8 and face recognition.
- Publication2024IEEE ICCCI 2024
Responsive Design using HTML and Tailwind CSS
Adaptive layouts, accessibility, and cross device user experience.
- Talk2026DTS AI Symposium
Transforming Learning and Teaching Assistance with BearcatGPT AI Agents
Presented multi agent tutoring workflows and retrieval design to university leadership and faculty.
- Recognition2024MakeUC 2024
Best Security Hack
Awarded for Secure Sight, the real time weapon detection platform.
Source - Workshop2025RAISE AI Summit
Responsible AI workshops
Guided 150 plus high school students through responsible AI workshops.
- Workshop2025University of Cincinnati
GenAI and Copilot enablement
Trained 200 plus faculty and staff on generative AI and Microsoft Copilot.
Technical stack
The stack I actually ship with
Grouped by what they do, not by alphabetical order. Production tools first. Things I am still exploring are clearly labeled.
AI and LLM systems
Production retrieval, agents, grounding, and prompt design across Azure AI and frontier models.
Orchestration and automation
No code and low code pipelines that move real work between systems on a schedule or trigger.
Backend engineering
Python and Node services that put guardrails, auth, and clean contracts in front of models and tools.
Frontend engineering
Type safe, accessible, recruiter ready interfaces for AI products and dashboards.
Data and ML
Classical ML, computer vision, and document extraction work that supports the AI stack.
Cloud and DevOps
Azure first, with AWS and Vercel for shipping, plus the boring parts that keep things alive in production.
Data engineering
Pipelines that move and shape data before models ever see it.
Practices
How I work day to day, plus the responsible AI guardrails that ship with every product.
Currently exploring
Tools I am actively learning and prototyping with. Not yet shipped in production, and labeled as such.
Currently exploring
Frameworks I am actively learning and prototyping with, but have not yet shipped in production.
Selected writing
Engineering notes in progress
Drafts on the architecture, tradeoffs, and edge cases behind the work.
What I learned building grounded RAG on Azure AI Search
Chunking, hybrid retrieval, re ranking, citations, and the latency tuning that cut response time by 30 percent.
Why I skipped a vector database for meeting intelligence
How a flat SharePoint list beat a vector index when the access pattern was per person and date range, not semantic similarity.
Bridging binary enterprise documents to LLM tool calls
The TAMALE extraction gateway pattern: streaming PDFs, DOCX, XLSX, and PPTX into structured JSON envelopes safely.
Let's build something
Open to AI Engineer, Applied AI Engineer, Full Stack AI Engineer, and AI ML roles.
The fastest way to reach me is email. I reply within a business day.