Aakash Madabhushi
AI Engineer I build applications on language models that look up evidence before they answer and ask a person before they act.
0.87 s median time to a plain-English, evidence-grounded explanation of an alert
0.90 F1 flagging suspicious machine pairs, 0.94 precision and 0.87 recall
14× faster model chosen for real-time robot control, 18 ms against 251 ms a decision Gave security analysts a plain-English reason for every alert in 0.87 s at the median: the system retrieves network context and threat intelligence from a Pinecone index, and Llama 3.3 explains the flag from that evidence. Held every agent to one rule, that a person approves anything irreversible. InvoSense reads invoices with a vision model, rates its confidence in each field, and cannot trigger a payment until someone signs off. Shipped a LangGraph recruiting assistant that goes from résumé to drafted outreach and stops before sending, and a reading assistant whose model, Gemini Nano, runs on the user's own device so nothing they read leaves it. Worked below the frameworks too: a GPT-2 and a byte-pair tokenizer written from scratch, and a Phi-3 fine-tune. 7 projects in this area Every link goes to code or data you can inspect yourself.
Sep 2026 · Hackathon
Tell: Catching a Payments Agent Being Talked Into Theft Reads a payments agent's hidden state before each decision and hands risky ones to a safer copy of itself, on one local GPU.
Jun 2026 · Personal project
InvoSense: Invoice Processing with a Human Approval Gate Reads an invoice, admits which fields it is unsure about, and cannot pay a cent until a person has approved it.
Mar 2026 · Hackathon
AgentForge, a Self-Healing API Monitor Notices an API has broken, works out why, and fixes it without waking anyone up. Built in five hours at a hackathon.
Spring 2026 · Coursework
LinkedIn Simulation with an AI Recruiting Assistant Takes a recruiter from a pile of résumés to a shortlist without slowing the page, and asks before contacting anyone.
Jul 2026 to present · Personal project
Reader, an Attention-Management Browser Extension Tells you where your attention actually went today, and keeps every page you read on your own machine instead of on a server.
Feb to May 2026 · Team coursework
Explainable Network Threat Detection Finds the suspicious pair of machines among 27.1M network connections, and tells the analyst in plain English why.
Nov 2025 · Personal project
Semantic Article Search Pipeline Finds the article you meant rather than the words you typed, by indexing each one by meaning on a nightly schedule.
Select a project for its figures, stack and links.
× Tell: Catching a Payments Agent Being Talked Into Theft SJSU hackathon, run on an HP ZGX Nano · Sep 2026 · AI Engineering, ML Engineering
Reads a payments agent's hidden state before each decision and hands risky ones to a safer copy of itself, on one local GPU.
Results
0.841 AUROC on held-out invoice attacks, 400 records from 50 documents
17 GB GPU memory to serve Qwen3-8B and its adapter on one GB10
76 / 76 clean invoices still paid correctly with routing on Built with Python · PyTorch · CUDA · Qwen3-8B · Hugging Face Transformers · LoRA · NVIDIA GB10 · Tesseract · SQLite
× InvoSense: Invoice Processing with a Human Approval Gate Personal project · Jun 2026 · AI Engineering, ML Engineering
Reads an invoice, admits which fields it is unsure about, and cannot pay a cent until a person has approved it.
Built with Python · LangGraph · Groq · Llama 3.2 Vision · Llama 3.1 · ChromaDB · React
× AgentForge, a Self-Healing API Monitor SJSU Applied Data Science Hackathon 2026 · Mar 2026 · AI Engineering, Software Engineering
Notices an API has broken, works out why, and fixes it without waking anyone up. Built in five hours at a hackathon.
Built with Python · FastAPI · LLM · Railtracks · React
× LinkedIn Simulation with an AI Recruiting Assistant SJSU DATA 236, Distributed Systems · Spring 2026 · Software Engineering, AI Engineering, Data Engineering
Takes a recruiter from a pile of résumés to a shortlist without slowing the page, and asks before contacting anyone.
Results
6× faster repeat profile loads, under 100 concurrent users
17 containers across 5 service groups Built with FastAPI · Kafka · LangGraph · React · MySQL · MongoDB · Redis · Docker
× Reader, an Attention-Management Browser Extension Personal project · Jul 2026 to present · Software Engineering, AI Engineering
Tells you where your attention actually went today, and keeps every page you read on your own machine instead of on a server.
Built with TypeScript · React · Vite · Chrome MV3 · SwiftUI · Cloud Firestore · Gemini Nano
× Explainable Network Threat Detection SJSU capstone, team project · Feb to May 2026 · ML Engineering, AI Engineering, Data Engineering
Finds the suspicious pair of machines among 27.1M network connections, and tells the analyst in plain English why.
Results
0.90 F1 flagging suspicious machine pairs (0.94 precision)
27.1M Zeek network flows aggregated into the graph
0.87 s median end-to-end latency per retrieval-grounded explanation Built with PyTorch Geometric · Airflow · dbt · Snowflake · Pinecone · Llama 3.3
× Semantic Article Search Pipeline Personal project · Nov 2025 · Data Engineering, AI Engineering
Finds the article you meant rather than the words you typed, by indexing each one by meaning on a nightly schedule.
Built with Airflow · Sentence Transformers · Pinecone · Docker
Hiring for AI engineering? I'm available from December 2026 and open to relocating. Email is the fastest way to reach me.
Hiring for something else?
