AI Engineer ยท Researcher ยท Author of QEMA-G
VP of R&D @ AI Skunkworks, Northeastern University ยท Boston, MA
I build AI systems that survive contact with reality โ then publish the research to prove it. Author of QEMA-G, a quantum-enhanced memory framework for graph AI (Zenodo ยท TechRxiv), and VP of R&D at AI Skunkworks, Northeastern's graduate AI research org. Four years of production engineering before grad school; now 30+ open-source projects deep into the space where multi-agent systems, RAG, and reinforcement learning converge.
- ๐ MS in Information Systems @ Northeastern University โ GPA 3.90/4.0
- ๐ 1st Place, MGEN (now SEIS) Hackathon 2025 โ MSL Practice Gym, an AI practice platform for Medical Science Liaisons built in 48 hours for the DNATE challenge
- โ๏ธ QEMA-G: Zenodo ยท TechRxiv ยท plain-English explainer ยท validation code
- ๐ Author of the Ideas You Can Picture book series โ on Amazon
- ๐ก๏ธ New: VeriFact โ open-source credibility engine for news & social media: source reputation, claim verification, fact-check lookup, corroboration & media provenance (CLI + API + web UI + Chrome extension)
- ๐ผ Open to AI/ML Engineering & Applied Research roles โ aravindbalaji.com
| Project | Description | Tech |
|---|---|---|
| agrocortex | The open-source AI cortex for agriculture: LLM benchmark, plant-disease vision, satellite field analytics, and a citation-grounded AI agronomist. | Python |
| Cerebro | CEREBRO โ AI Research Assistant | Python |
| ch09-grounding-agents-in-evidence | Production case study of Magid's Collaborator newsroom RAG platform โ why retrieval grounding is an epistemological constraint for agentic AI, with figures and notebook. | HTML |
| civisynth | Open political intelligence platform: narrative tracking, RAG fact-checking with evals, legislative intelligence, and election forecasting. | Python |
| CodeSentinel-v1 | CodeSentinel v1 โ Multi-Agent AI System for Intelligent Code Review and Vulnerability Detection (original version, split from Projects monorepo) | โ |
| CodeSentinel-v2 | CodeSentinel v2 โ Multi-Agent AI System for Intelligent Code Review and Vulnerability Detection | Python |
| Foglight | All-seeing internet intelligence layer for agents and humans โ pluggable channels, persistent memory, temporal knowledge graph, monitoring and cited synthesis. | Python |
| LLM-Agents-Deep-Q-Learning-Atari | LLM Agents & Deep Q-Learning with Atari Games | Jupyter Notebook |
| medigraph-ai | MediGraph AI โ offline-first clinical knowledge-graph & decision-support platform with FHIR/HL7/C-CDA connectors for US + India | Python |
| MediGraphAI | LLM-Enabled Healthcare Knowledge Graph Platform | Python |
| omnicanon | Citation-verified AI infrastructure for sacred texts - multi-faith corpus, zero-hallucination RAG, TheoBench LLM benchmark, and canon-aware OCR. | Python |
| QueryCraft | Fine-tuning an LLM - QueryCraft | Jupyter Notebook |
| RL_Codesentinel | Reinforcement learning for agentic code review โ multi-agent reviewers that learn from developer feedback (contextual bandits + policy gradient, 200-episode ร 5-seed evaluation). | Python |
| verifact | Open-source credibility engine for news, web content and social media - provenance, claim verification, source reputation and manipulation signals. | Python |
| verimathix | Verified-math toolkit: step-level solution verification, error localization, Lean 4 autoformalization, numeric kernels, and a step-grading benchmark. | Python |
| WardSight | Open-source intelligence fusion platform with a citation-grounded AI analyst copilot (ADS-B/AIS/GDELT fusion, entity resolution, anomaly detection, grounded RAG). | Python |
| Project | Description | Tech |
|---|---|---|
| everycam | Every camera is a robot teacher: privacy-first, robot-ready embodied data from everyday cameras. | Python |
| RuView | ฯ RuView: WiFi DensePose turns commodity WiFi signals into real-time human pose estimation, vital sign monitoring, and presence detection โ all without a single pixel of video. | โ |
| Smart-Trolley | Edge-AI smart shopping trolley on Raspberry Pi โ CNN-based produce recognition (VGG16 โ MobileNetV3, TensorFlow Lite INT8), load-cell weighing, and real-time Firebase billing & inventory. | Python |
| Project | Description | Tech |
|---|---|---|
| agrocortex | The open-source AI cortex for agriculture: LLM benchmark, plant-disease vision, satellite field analytics, and a citation-grounded AI agronomist. | Python |
| everycam | Every camera is a robot teacher: privacy-first, robot-ready embodied data from everyday cameras. | Python |
| RuView | ฯ RuView: WiFi DensePose turns commodity WiFi signals into real-time human pose estimation, vital sign monitoring, and presence detection โ all without a single pixel of video. | โ |
| Smart-Trolley | Edge-AI smart shopping trolley on Raspberry Pi โ CNN-based produce recognition (VGG16 โ MobileNetV3, TensorFlow Lite INT8), load-cell weighing, and real-time Firebase billing & inventory. | Python |
| Project | Description | Tech |
|---|---|---|
| alchemind | Invent, predict, and validate novel chemical compounds โ an end-to-end de novo molecular discovery platform (RDKit + genetic generation + ML property prediction + FastAPI). | Python |
| biopolyglot | BioPolyglot: one model, many biological languages โ a modular multi-modal ML platform unifying protein variant-effect, molecular property, genomic, and single-cell prediction behind one interface. | Python |
| carbonoscope | Open climate intelligence platform: auditable GHG accounting, Climate TRACE analytics, grid carbon forecasting, disclosure parsing, climate risk. | Python |
| LLM-Agents-Deep-Q-Learning-Atari | LLM Agents & Deep Q-Learning with Atari Games | Jupyter Notebook |
| physweave | Weaving physics simulation and machine learning into one modular toolkit: differentiable simulation, neural PDE surrogates, stochastic finance, quantum circuits, and molecular dynamics. | Python |
| Projects | Academic Projects | Python |
| QueryCraft | Fine-tuning an LLM - QueryCraft | Jupyter Notebook |
| RL_Codesentinel | Reinforcement learning for agentic code review โ multi-agent reviewers that learn from developer feedback (contextual bandits + policy gradient, 200-episode ร 5-seed evaluation). | Python |
| voltiq | โก EV Battery Intelligence Platform โ CAN telemetry ingestion, ML anomaly detection, SOH estimation & RUL prediction | Python |
| Project | Description | Tech |
|---|---|---|
| voltiq | โก EV Battery Intelligence Platform โ CAN telemetry ingestion, ML anomaly detection, SOH estimation & RUL prediction | Python |
| Project | Description | Tech |
|---|---|---|
| carbonoscope | Open climate intelligence platform: auditable GHG accounting, Climate TRACE analytics, grid carbon forecasting, disclosure parsing, climate risk. | Python |
| Project | Description | Tech |
|---|---|---|
| agrocortex | The open-source AI cortex for agriculture: LLM benchmark, plant-disease vision, satellite field analytics, and a citation-grounded AI agronomist. | Python |
| Project | Description | Tech |
|---|---|---|
| biopolyglot | BioPolyglot: one model, many biological languages โ a modular multi-modal ML platform unifying protein variant-effect, molecular property, genomic, and single-cell prediction behind one interface. | Python |
| CareHub | Hospital Management System | HTML |
| ClearAuth | Open-source CMS-0057-F prior authorization platform โ FHIR-native PAS intake, auditable adjudication, SLA enforcement, CMS PA metrics | Java |
| medigraph-ai | MediGraph AI โ offline-first clinical knowledge-graph & decision-support platform with FHIR/HL7/C-CDA connectors for US + India | Python |
| MediGraphAI | LLM-Enabled Healthcare Knowledge Graph Platform | Python |
| MGEN-Hackathon-2025 | ๐ 1st Place โ MSL Practice Gym, MGEN (now SEIS) Hackathon 2025, Northeastern University. AI practice platform for Medical Science Liaisons. | โ |
| Project | Description | Tech |
|---|---|---|
| ClearAuth | Open-source CMS-0057-F prior authorization platform โ FHIR-native PAS intake, auditable adjudication, SLA enforcement, CMS PA metrics | Java |
| quantumteller | The open-source ATM stack โ from COBOL core to quantum-safe crypto. ISO 8583 switch (Go) + COBOL core banking + roadmap to ML fraud scoring and post-quantum crypto. | Go |
| Project | Description | Tech |
|---|---|---|
| civisynth | Open political intelligence platform: narrative tracking, RAG fact-checking with evals, legislative intelligence, and election forecasting. | Python |
| verifact | Open-source credibility engine for news, web content and social media - provenance, claim verification, source reputation and manipulation signals. | Python |
| Project | Description | Tech |
|---|---|---|
| orbistra | Open-source space operations platform: parallel SGP4 propagation, full-catalog conjunction screening, collision risk, 3-D ops console | Rust |
| Project | Description | Tech |
|---|---|---|
| CodeSentinel-v1 | CodeSentinel v1 โ Multi-Agent AI System for Intelligent Code Review and Vulnerability Detection (original version, split from Projects monorepo) | โ |
| CodeSentinel-v2 | CodeSentinel v2 โ Multi-Agent AI System for Intelligent Code Review and Vulnerability Detection | Python |
| Foglight | All-seeing internet intelligence layer for agents and humans โ pluggable channels, persistent memory, temporal knowledge graph, monitoring and cited synthesis. | Python |
| RL_Codesentinel | Reinforcement learning for agentic code review โ multi-agent reviewers that learn from developer feedback (contextual bandits + policy gradient, 200-episode ร 5-seed evaluation). | Python |
| verifact | Open-source credibility engine for news, web content and social media - provenance, claim verification, source reputation and manipulation signals. | Python |
| WardSight | Open-source intelligence fusion platform with a citation-grounded AI analyst copilot (ADS-B/AIS/GDELT fusion, entity resolution, anomaly detection, grounded RAG). | Python |
| Project | Description | Tech |
|---|---|---|
| qemag-validation | Qiskit simulation code reproducing all numerical results of the QEMA-G paper โ quantum message-passing on graphs under noiseless and IBM-Brisbane noise models. | Python |
| quantumteller | The open-source ATM stack โ from COBOL core to quantum-safe crypto. ISO 8583 switch (Go) + COBOL core banking + roadmap to ML fraud scoring and post-quantum crypto. | Go |
| Project | Description | Tech |
|---|---|---|
| alchemind | Invent, predict, and validate novel chemical compounds โ an end-to-end de novo molecular discovery platform (RDKit + genetic generation + ML property prediction + FastAPI). | Python |
| biopolyglot | BioPolyglot: one model, many biological languages โ a modular multi-modal ML platform unifying protein variant-effect, molecular property, genomic, and single-cell prediction behind one interface. | Python |
| orbistra | Open-source space operations platform: parallel SGP4 propagation, full-catalog conjunction screening, collision risk, 3-D ops console | Rust |
| physweave | Weaving physics simulation and machine learning into one modular toolkit: differentiable simulation, neural PDE surrogates, stochastic finance, quantum circuits, and molecular dynamics. | Python |
| verimathix | Verified-math toolkit: step-level solution verification, error localization, Lean 4 autoformalization, numeric kernels, and a step-grading benchmark. | Python |
| Project | Description | Tech |
|---|---|---|
| ch09-grounding-agents-in-evidence | Production case study of Magid's Collaborator newsroom RAG platform โ why retrieval grounding is an epistemological constraint for agentic AI, with figures and notebook. | HTML |
| ideas-you-can-picture | Official website for the Ideas You Can Picture book series by Aravind Balaji | HTML |
| omnicanon | Citation-verified AI infrastructure for sacred texts - multi-faith corpus, zero-hallucination RAG, TheoBench LLM benchmark, and canon-aware OCR. | Python |
| Project | Description | Tech |
|---|---|---|
| quantumteller | The open-source ATM stack โ from COBOL core to quantum-safe crypto. ISO 8583 switch (Go) + COBOL core banking + roadmap to ML fraud scoring and post-quantum crypto. | Go |
| robust-data-processor | Fault-tolerant multi-tenant data ingestion pipeline on GCP โ Cloud Run API, Pub/Sub buffering, worker retries with dead-lettering, and load testing. | Python |
| ServiQue | ServiQue v1 โ service booking platform (Java Swing + DB4O); rewritten as servique-v2 | โ |
| servique-v2 | ServiQue v2 โ production-ready rewrite of the ServiQue service booking platform (v1: Java Swing/DB4O). Node+TypeScript+Express+Prisma API, React SPA, JWT auth with RBAC, tests, Docker, CI. | TypeScript |
| Project | Description | Tech |
|---|---|---|
| cheese-chase-arcade | ๐ง Ten arcade games + a 10-chapter story mode in a single HTML file. Open-source, zero dependencies. | HTML |
| Project | Description | Tech |
|---|---|---|
| aravindb98.github.io | Profile page โ live categorized projects | HTML |
| ideas-you-can-picture | Official website for the Ideas You Can Picture book series by Aravind Balaji | HTML |
| ServiQue | ServiQue v1 โ service booking platform (Java Swing + DB4O); rewritten as servique-v2 | โ |
| servique-v2 | ServiQue v2 โ production-ready rewrite of the ServiQue service booking platform (v1: Java Swing/DB4O). Node+TypeScript+Express+Prisma API, React SPA, JWT auth with RBAC, tests, Docker, CI. | TypeScript |
๐ This list updates automatically every day via GitHub Actions โ repos made public appear here on the next run.
Python PyTorch LangChain LangGraph CrewAI Hugging Face Anthropic API OpenAI API Neo4j ChromaDB Snowflake AWS Azure GCP React Node.js FastAPI Docker PennyLane
๐ซ Reach me: balaji.ara@northeastern.edu ยท Schedule a call