# Aditya Mhaske

> Software engineer building retrieval and multi-agent AI systems, from prototype to production: hybrid search, multi-tenant vector databases, LLM evaluation, and production ML pipelines. 5 IEEE publications, 2 patents.

- Currently at Leadec, building the retrieval layer for an industrial AI assistant.
- Multi-tenant vector search, hybrid keyword + semantic retrieval, 30x faster p99 lookups.
- Previously: 1M+ record embedding pipelines, Airflow ETL on AWS, fine-tuned BERT over 100M+ rows.

## Pages

- [Overview](https://adityamhaske.com): Software engineer building retrieval and multi-agent AI systems. Multi-tenant vector search for 100+ tenants and a 30x p99 latency cut at Leadec; 5 IEEE papers, 2 patents.
- [Experience](https://adityamhaske.com/experience): Work history: multi-tenant vector search and hybrid retrieval at Leadec, 1M+ record embedding pipelines at BMR Infotek, fine-tuned BERT research at Indiana University.
- [Projects](https://adityamhaske.com/projects): AI engineering projects: a multi-agent research assistant, Agent Arena for LLM evaluation, an AI co-pilot for SaaS analytics, multimodal deep learning and computer vision.
- [Publications & Patents](https://adityamhaske.com/research): Five peer-reviewed IEEE papers and two filed patents: EEG-based depression detection, autonomous braking, cardiovascular disease prediction, smart glasses for monocular vision.
- [Articles](https://adityamhaske.com/articles): Long-form engineering writing: multi-tenant isolation for vector search, a teardown of an agentic AI coding workflow, and earlier essays on India's pharmaceutical supply chain.
- [Contact](https://adityamhaske.com/contact): Get in touch with Aditya Mhaske about retrieval systems, multi-agent infrastructure, and applied LLM work.

## Articles

- [I Don't Prompt. I Build the Machine That Prompts.](https://adityamhaske.com/articles/workflow): An interactive teardown of my AI engineering stack: 43 on-demand skills, deterministic hooks that can block a tool call, 9 MCP servers, a seven-agent fleet running in parallel worktrees, and a context window run as a budget.
- [Isolation Is Four Dials. Everyone Ships It as One Switch.](https://adityamhaske.com/articles/multi-tenancy): A field study of multi-tenant isolation for retrieval systems: why the relational silo/pool/bridge playbook misprices vector stores, how a tenant filter silently degrades recall over a shared proximity graph, and the four-axis model — storage, compute, enforcement, operations — behind a 100+ tenant deployment that cut resident memory 80%. Interactive cost model, four figures, 15 primary sources.
- [Solutions in Achieving Self-Reliance in India's API Industry](https://adityamhaske.medium.com/solutions-for-dependency-of-indian-pharmaceutical-industry-on-china-for-apis-9b5c16fd909): Discusses the dependency of the Indian pharmaceutical industry on China for APIs and potential solutions.
- [Dependency of Indian Pharmaceutical Industry on China for APIs](https://adityamhaske.medium.com/dependency-of-indian-pharmaceutical-industry-on-china-for-apis-e1fc14936b44): Highlights growth and export figures of the Indian pharmaceutical sector and increasing competition from China.

## Projects

- [AI Co-Pilot for SaaS Analytics Platform](https://github.com/adityamhaske/AI-Co-Pilot-for-SaaS-Analytics-Platform): Built an AI co-pilot enabling natural-language access to a SaaS analytics dashboard via FastAPI and Anthropic function calling, mapping user queries to structured backend calls.
- [Multi-Agent Research Assistant](https://github.com/adityamhaske/Multi-Agent-Research-Assistant): Built a multi-agent research platform orchestrating planner, researcher, critic, and synthesis agents with tool-calling and shared state management to automate multi-step research workflows.
- [Computer Use Automation](https://github.com/adityamhaske/computer-use-automation): Built a computer-use system where an LLM works out how to reach a goal in a legacy application with no API exactly once; the run is compiled into a typed, versioned capability artifact.
- [Agent Arena](https://github.com/adityamhaske/agent-arena): Engineered an evaluation system for LLM models providing a universal arena to determine model selection based on accuracy, cost, and latency, alongside a multi-agent study framework.
- [IndusAI](https://github.com/adityamhaske/IndusAI): Engineered an industrial AI assistant for PLC fault log analysis; migrated to Google Cloud (Gemini Flash, Qdrant, Cloud Run, Firebase)
- [AffectiSense](https://github.com/adityamhaske/AffectiSense): Engineered a multimodal mental health assessment platform fusing neurophysiological (EEG), vocal biomarker (audio), and behavioral (facial expressions) signals for depression screening.
- [Multimodal Depression Analysis](https://github.com/adityamhaske/Multimodal_Depression_Analysis): Built a real-time system recognizing depression by extracting features from facial expressions, EEG signals, and speech modules using multimodal deep learning.
- [Object Detection & Localization](https://github.com/adityamhaske/Object-Detection-and-Localization): Trained YOLOv3 on the COCO dataset for real-time object detection and localization on road scenes; related to autonomous vehicle patents and deployed in a braking system prototype
- [Stock Index & GDP Forecasting](https://github.com/adityamhaske/Financial-Econometrics): Developed an ARIMA model built with MATLAB's Econometrics Toolbox to forecast GDP and Stock Indexes of 10 countries.
- [Business Location Recommendation System](https://github.com/adityamhaske/Business-Location-Recommendation-System): Architected a data analysis and ML pipeline to find optimal business locations using Wikipedia data and Foursquare API for Toronto neighborhoods; features Folium mapping, K-Means clustering, and geospatial visualization

## Research

- [EEG Based Depression Detection Using Ensemble Approach](https://ieeexplore.ieee.org/abstract/document/10201102): International Conference on Applied Intelligence and Sustainable Computing
- [Road Conditions Indication and Autonomous Braking System](https://ieeexplore.ieee.org/document/9315862): IEEE 3rd International Conference on Intelligent Sustainable Systems
- [Cardiovascular Disease Prediction Using Machine Learning Models](https://doi.org/10.1109/PuneCon50868.2020.9362367): IEEE Pune Section International Conference
- [Smart Glasses to Assist Monocular Vision to Estimate Depth](https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=9298311): IEEE International Conference for Innovation in Technology 2020, Bangalore
- [A Survey: Feature Extraction Techniques And ML Models For Depression Analysis](https://doi.org/10.1109/ICCCNT51525.2021.9579727): 12th Intl. Conf. on Computing, Communication, and Networking Technologies
- Smart Glasses for Monocular Vision: patent, IN 202021034153 · Filed Aug 10, 2020
- Road Conditions and Obstacle Indication System: patent, IN 202021009028 · Filed Mar 3, 2020

## Profiles

- [LinkedIn](https://www.linkedin.com/in/adityamhaske/)
- [GitHub](https://github.com/adityamhaske)
- [Google Scholar](https://scholar.google.com/citations?user=klyUNDAAAAAJ&hl=en)
- [Kaggle](https://www.kaggle.com/adityamhaske)
- [Medium](https://adityamhaske.medium.com/)

## Optional

- [Full profile](https://adityamhaske.com/llms-full.txt): complete work history, education, projects, publications, patents, certifications, and articles as one Markdown file