Most engineers ship code. I ship research that ships back. Author of QEMA-G, a quantum‑enhanced memory framework for graph AI, published on TechRxiv and Zenodo. VP of R&D at AI Skunkworks, the graduate research org I rebuilt from the ground up and got formally recognised by Northeastern's Graduate Student Government. Four years of production engineering before grad school, now building the layer where multi-agent systems, RAG, and reinforcement learning converge into LLMs that actually do useful work in the real-world AI ecosystem.
Two things in flight as of Spring 2026: research leadership and graduate work converging on the same problem, how to make agentic systems that actually work in production.
Leading R&D for a graduate research organization recognised by Northeastern University Graduate Student Government, advised by Prof. Nik Bear Brown. Recruiting and mentoring graduate researchers across multi-agent systems, RAG, and applied LLM research.
Author of QEMA-G, a theoretical framework integrating quantum memory with graph neural networks to address the memory-wall problem. Co-authored with Prof. Nik Bear Brown. Published on TechRxiv and Zenodo.
Multi-agent systems, retrieval pipelines, fine-tuned models, and production data platforms. Each one shipped, tested, or graded, not a tutorial follow-along.
940-line custom multi-agent system querying 16 academic databases in parallel through a CrewAI + n8n architecture, surfaced through a Streamlit UI. Built for INFO 7375 (Building Agentic Systems) under Prof. Nik Bear Brown.
Theoretical framework integrating quantum memory architectures with graph neural networks to attack the memory-wall problem in large-scale GNN training. Author: Aravind Balaji · Co-author: Prof. Nik Bear Brown. Published on TechRxiv and Zenodo. Companion executive briefing deck built for C-suite audiences with an applied pilot roadmap.
CodeSentinel is a LangGraph-based multi-agent code review system. Multiple specialised reviewer agents collaborate over a shared graph state to surface bugs, style issues, and security concerns. RL-CodeSentinel extends it by layering UCB-1 Contextual Bandit and REINFORCE Policy Gradient on top, so reviewer-agent selection improves with feedback rather than staying static. Together they show how RL signal can be added to existing agentic pipelines without rebuilding orchestration.
Fine-tuned TinyLlama-1.1B with QLoRA on the Spider dataset for natural-language-to-SQL generation. Achieved 100% improvement on SELECT-clause generation against the base model. Shipped with a Gradio demo and full RAG retrieval layer over schema metadata.
Healthcare reasoning system combining a Neo4j knowledge graph with LLM-driven retrieval over Snowflake-hosted patient data. End-to-end RAG architecture: graph-aware retrieval, structured reasoning, and natural-language explanations of clinical relationships.
Won 1st place at the MGEN Hackathon 2025 leading a 4-person team. AI conversation simulator for Medical Science Liaisons, with role-played physician interactions and structured feedback. Built end-to-end in 48 hours.
10-function career-assistance toolkit built on the Anthropic API: JD analyzer, resume builder, cover letter generator, and application tracker. A working example of practical LLM tool-use orchestration on a single agent.
Contributed Chapter 6, "Grounding Agents in Evidence," to Prof. Nik Bear Brown's textbook on agentic systems. The chapter covers retrieval-augmented generation as an evidence layer for LLM agents, evaluation methodology for citation faithfulness, and design patterns for grounding agentic outputs in source material rather than free-form generation.
Four years across healthcare technology, cloud migrations, and SaaS reliability before grad school. Shipping in production, not just classrooms.
Deep on AI/ML and Python; comfortable across the data, cloud, and full-stack layers I need to ship end-to-end systems.
Three degrees across electronics, software engineering, and information systems, with a particular emphasis on the layer where AI meets real systems.
Narrative nonfiction on AI engineering, quantum computing, and the human side of the systems we build. Published on Substack.
The mathematics of signal processing is the mathematics of attention mechanisms. The engineers who can bridge hardware and software are the ones the AI era needs most.
A plain-English walkthrough of the QEMA-G framework. Quantum-enhanced memory for graph AI, told as a story rather than a paper.
What custom AI silicon actually buys you, where Trainium fits against H100s, and why the chip wars have only just started.
Books that have shaped how I think about systems, attention, and engineering. Click through for reviews on Substack.
System 1 and System 2 map almost directly onto fast inference vs. slow reasoning in modern AI.
The best-designed systems solve a problem you couldn't articulate until the solution was in your hands.
Reading isn't just consumption. It's the act of rebuilding your attention span one page at a time.
The gap between a model in a notebook and a model in production is where real engineering lives.
The bridge between quantum theory and real circuits. Essential for anyone building QEMA-G style systems.
Every moment in business happens only once. The next Bill Gates will not build an operating system.
The rhythm behind the work, what fuels long research sessions and late-night builds.
Most of what plays during deep work is something I made myself. I produce independent music using AI tools, composing, prompting, iterating, and curating until the track sounds the way I imagined it before I started.
It's the same loop as engineering: idea → prototype → critique → ship. Just with melody instead of code. Hit the Suno link for the full track.
Listen on SunoMost candidates pick a lane. Production engineering or research. Models or systems. Theory or shipping. I'm fluent in all three, which is the rarer profile, and the one that makes AI initiatives actually move.
Open to: AI / ML Engineering · Applied Research · Multi-Agent Systems · Data & ML Platforms · Full-Stack with an AI layer.
When: Co-op or Internship starting Summer 2026 for up to 8 months, then full-time starting December 2026. No visa sponsorship required.
Where: In-person or remote, anywhere in the United States. Reply window: within 24 hours, every time.
I'm open to all roles across the United States and anywhere in the world. If you're hiring for a position where research-grade thinking has to ship in production, let's talk.