# Ainesh Chatterjee

AI Student Researcher & Systems Engineer
I code often and sleep occasionally. Agents are graphs.

## About

Undergraduate researcher at the University of Maryland, pursuing a dual BS in Computer Science (Machine Learning) and Mathematics. My work focuses on agentic AI systems, deep reinforcement learning, and building production-ready ML infrastructure. Currently developing autonomous agents at Tilli Software using MCP (Model Context Protocol) and advanced context engineering. Research interests span GAIfO-based imitation learning, non-Euclidean crowd simulation, and quantum-safe cryptography.

### Education
**University of Maryland - College Park**
- Dual BS in Computer Science (Machine Learning) and Mathematics
- Graduation: December 2025
- Honors: University, CS Departmental Honors, BS/MS, Dean's List

### Focus Areas

**Machine Learning & AI**
Specializing in deep reinforcement learning, imitation learning (GAIfO), and large language models for real-world tasks.

**Systems Engineering**
Optimizing DevOps pipelines, container-based infrastructures, and multi-agent simulation frameworks at scale.

**Quantum Computing**
Exploring lattice-based cryptography and quantum-safe protocols, ensuring security in the post-quantum era.

### Coursework
- **AI/ML**: Graduate NLP, HRI/Embodied AI, Computer Vision, Intro to: AI, ML, Data Science, Multimodal Deep Learning (Intro)
- **Math**: Calc III, Advanced Linear Algebra, Differential Equations, Advanced Calculus, Abstract Algebra, Mathematical Finance: Derivatives & Stochastic Models, Transform Methods, Numerical Analysis
- **CS**: Quantum Computing, Algorithms, Data Structures, Computer Systems, Object-Oriented Programming, Organization of Languages
- **Statistics**: Applied Probability & Statistics, Probability Theory

## Research

My research focuses on bridging theoretical insights in ML, mathematics, and quantum-safe protocols with tangible real-world applications. I strive to build robust, efficient systems and algorithms that address modern challenges in security, large-scale data, and multi-agent autonomy.

### Current Research

**Demosthenes: Training LLMs for Legal Argumentation**
Developing a framework for training and evaluating LLMs on legal argumentation tasks using SoTA approaches.
Technologies: Python, PyTorch, Transformers, NLP

## Publications

My research work spans computational geometry and artificial intelligence, with a focus on practical applications and optimally leveraging theoretical foundations.

### Ipelets for the Convex Polygonal Geometry
- **Venue**: SoCG 2024 2024
- **Authors**: Ainesh Chatterjee, Nithin Parepally, Auguste H. Gezalyan, Hongyang Du, Sukrit Mangla, Kenny Wu, Sarah Hwang, David M. Mount
- **Type**: conference
- **DOI**: 10.4230/LIPIcs.SoCG.2024.92
- **URL**: https://doi.org/10.4230/LIPIcs.SoCG.2024.92
- **Code**: https://github.com/ain3sh/umd_ipelets

**Abstract**: In this media submission, we showcase a collection of new Ipelets that construct a variety of geometric objects based on polygonal geometries.

### AgreeMate: Teaching LLMs to Haggle
- **Venue**: arXiv 2024
- **Authors**: Ainesh Chatterjee, Samuel Miller, Nithin Parepally
- **Type**: preprint
- **URL**: https://arxiv.org/abs/2412.18690
- **Code**: https://github.com/ain3sh/agreemate

**Abstract**: We introduce AgreeMate, a framework for training Large Language Models (LLMs) to perform strategic price negotiations through natural language.

## Projects

A selection of my primary projects that highlight diverse skills—from AI-driven applications and quantum-safe security tools to hackathon-winning solutions.

### Vizier
**Role**: Team Lead / ML Developer
**Status**: Completed

AI-powered personalized newsletter platform; MVP built for Bitcamp 2025
Test-time MoE agentic architecture improving context retrieval via document-expert LLMs

**Technologies**: Python, PyTorch, React, Docker
**Category**: ml

**Links**: [GitHub](https://github.com/ain3sh/vizier) | [Blog](/blog/projects/vizier)

### QSafe
**Role**: Solo Developer
**Status**: Active

Open-source Python/Rust password manager employing lattice-based encryption
Secured Docker container for key storage and CLI communication
Pursues end-to-end quantum-safe security with minimal overhead

**Technologies**: Python, Rust, Docker, Cryptography
**Category**: security

**Links**: [GitHub](https://github.com/ain3sh/qsafe)

### CoronaSafe
**Role**: Team Lead / Backend Developer
**Status**: Completed

Python/Flutter application providing real-time COVID-19 contraction risk ratings for global addresses
Incorporates foot traffic data, time-weighted algorithms, and real-time updates for predictive accuracy
Won the 2021 Congressional App Challenge (MD08)

**Technologies**: Python, Flutter, Plotly, Dask
**Category**: hackathon

**Achievements**:
- 2021 Congressional App Challenge Winner (MD08)

**Links**: [GitHub](https://github.com/ain3sh/coronasafe_flutter_app) | [Paper](https://www.congressionalappchallenge.us/21-md08/) | [Video](https://www.youtube.com/watch?v=64UmQ8RtvoY)

### Resourceful
**Role**: Team Lead / Backend Developer
**Status**: Completed

Helps students discover scholarships, courses, and internships via an NLP-driven search engine
Utilizes Python/Flutter, advanced text similarity, and dynamic scraping
Recipient of BlairHacks_5 Best Education Award (2022)

**Technologies**: Python, Flutter, NLTK, Selenium
**Category**: hackathon

**Achievements**:
- Best Education Award – BlairHacks_5 (2022)

**Links**: [GitHub](https://github.com/ain3sh/resourceful_flask_app) | [Demo](https://devpost.com/software/resourceful-7vydko)

### OptiTask
**Role**: Solo Developer
**Status**: Completed

Uses a 3D genetic algorithm to generate custom daily tasks from Canvas LMS data
Allows adjustable constraints and personal preferences for near-optimal scheduling
Originated from high school APEX research, now incorporating LLM expansions and hybrid diffusion approaches

**Technologies**: Python, Heroku, NumPy
**Category**: productivity

**Links**: [GitHub](https://github.com/ain3sh/optitask) | [Paper](https://quartz-suggestion-1ac.notion.site/33a62ea67d794e94a1a98f49625237c9)

## Experience

A chronicle of my work and research roles, spanning deep ML R&D, academic teaching, and DevOps optimization.

### AI Engineering Intern at Tilli Software
**July 2025 - Present** | Hybrid
*Edge:XDEX:Agent*

- Engineered: the Tilli Agent MVP (Pocketflow, Google Agent ADK) for utility customer web portals
- Developed: Scrape2MCP to convert arbitrary sites into structured API/browser actions; generated template-derived MCP servers with the Claude Code SDK
- Architected and optimized: a shared, multi-tenant MCP Super-Server as a tool store for user agents and Bedrock Agentcore deployment; instrumented automated performance logging for async analysis and release decisioning (increased cache-hit rate; reduced p50 latency and token cost)
- Leading: Tilli Agent launch (initial 150k+ users; planned rollout to ~3M)

### Student Researcher at University of Maryland CMNS
**September 2024 - June 2025** | College Park, MD
*Crowd Simulation*

- Explored: application of non-Euclidean geometries to crowd flow dynamics
- Applied: Transformer-based DNNs to crowd navigation, with a focus on natural language goal-direction

### Computer Science Intern at Johns Hopkins University Applied Physics Laboratory
**May 2024 - Aug 2024** | Laurel, MD
*Force Projection Sector: Ocean Systems & Engineering Group*

- Implemented: iteratively enhanced Generative Adversarial Imitation from Observation (GAIfO) agents (substantially outperforming baseline imitation models)
- Authored: critical literature reviews on GAIfO and Generative AI, providing (direct insights for future project strategies)
- Developed: an optimized GAIfO variant, using core-architectural insights from a literature review, which outperformed all prior versions over long timeframes
- Enhanced: GTRI's SCRIMMAGE mass-simulation framework with increased complexity and expert controller functionality
- Revamped: GitLab Continuous Integration pipelines, boosting speed and efficiency by (25% while addressing security vulnerabilities)
- Optimized: project-wide Docker Image, used across all repositories, reducing pipeline build times by (50% and increasing memory efficiency by 40%)
- Led: winning team for sector Intern Challenge in developing a secure, non-GPS intra-campus navigation prototype

### Lead Teaching Assistant at University of Maryland CMNS
**Spring 2024** | College Park, MD
*CMSC351H (Algorithms Honors)*

- Co-designed and graded: homeworks, exams, and lecture material for 38 honours students
- Conducted: weekly office hours, providing personalized guidance on advanced topics

### Research Intern at University of Maryland MIND Lab
**October 2023 - December 2024** | College Park, MD
*Breathing Analysis Project*

- Developed: an advanced visualization dashboard for efficient analysis of mass breath data
- Designed: dataset structures for visualization and feature extraction in future work
- Optimized: massive dataset-loading using Dask and multithreading by over (400%)
- Implemented: and evaluated supervised learning techniques for improved breath segmentation

### Bachelor of Science in Mathematics at University of Maryland - College Park
**February 2023 - Dec 2025** | College Park, MD
*Traditional Track*

- Honors: University; Dean's List
- Academic Performance: Maintained a competent GPA

### Bachelor of Science in Computer Science at University of Maryland - College Park
**Sep 2022 - Dec 2025** | College Park, MD
*Machine Learning Specialization*

- Honors: University, CS Departmental Honors; BS/MS; Dean's List
- Scholarships: Dean's Scholarship; National Merit
- Academic Performance: Maintained a competent GPA

### Support Staff/Host at Owen’s Tavern and Garden
**Jun 2022 - Aug 2022** | North Bethesda, MD
*Hospitality*

- Managed: Reservations, greeted patrons, processed orders both online and via phone, ensuring accurate fulfillment
- Collaborated: Effectively communicated with servers and managers to facilitate efficient seating arrangements
- Maintained: Cleaned and reset tables, upheld hygiene standards by polishing cutlery, mopping floors, and restocking bathrooms

### Tutor at Schoolhouse.world
**Jul 2021 - Sep 2022** | Remote
*Education*

- Tutored: Math courses and SAT Reading & Writing/Math, guiding students to academic success
- Developed: Personalized instruction plans to address individual student needs and learning styles
- Mentored: Inspired and motivated students to overcome academic challenges with confidence

### High School Diploma at Walter Johnson High School
**Sep 2018 - Jun 2022** | Bethesda, MD
*General Education*

- Leadership: Founder/President of the Coding Club; Co-Senior Captain of the Debate Team
- Honors: National Merit Scholarship Finalist; AP Scholar with Distinction; Eagle Scout
- Awards: Congressional App Challenge Winner (2021, MD08); BlairHacks_5 Best Education Award (2022)

## Skills

Specialized in machine learning frameworks, quantum computing tools, and high-performance computing systems. My tech stack is focused on research and development in AI, quantum computing, and distributed systems.

### Programming
- **Primary**: Python, C/C++, DevOps, Webhosting, Fullstack Development, API-creation, Design Paradigms
- **Secondary**: Java, Rust, Lua, MATLAB, Flutter/Dart, HTML5, CSS3, JavaScript, Assembly

### ML/AI
Transformers, Agentic LLMs, MCP, Context Engineering, Mechanistic Interpretability, DSPy, GEPA, RAG, Supervised/Unsupervised Learning, Deep RL, GANs

### Tools & Technologies
Git, GitHub/Lab, Linux, Bash, WSL2, PowerShell, FastAPI, Flask, RESTful, Postman, React, PyTorch, NumPy, Pandas, SciPy, Dask, scikit-learn, TensorBoard, Matplotlib, Seaborn, NLTK, spaCy, AWS EC2, AWS Fargate, AWS Lambda, AWS S3, AWS Bedrock, AWS SageMaker, Claude Code SDK, MCP, Google Agent ADK, Google Agent2Agent (A2A), Pocketflow, OpenAI API, HuggingFace, LiteLLM, PostgreSQL, NeonDB, Neo4j, SQL, Selenium, BeautifulSoup, ROS, LaTeX, Memory Profiler

### Finance
Brownian Motion, Black-Scholes, Arbitrage Pricing, Stochastic Calculus, Delta Hedging

### Soft Skills
First-Principles Problem Solving, Leadership, Technical Writing, Self-teaching, Iterative Experimentation

## Contact

- **Email**: contact@ain3sh.com
- **LinkedIn**: https://www.linkedin.com/in/ain3sh/
- **GitHub**: https://github.com/ain3sh/
- **Google Scholar**: https://scholar.google.com/citations?user=j87_K9EAAAAJ
- **Twitter/X**: https://x.com/ain3sh
- **Blog**: https://ain3sh.com/blog

## External Resources

- GitHub: https://github.com/ain3sh/
- LinkedIn: https://www.linkedin.com/in/ain3sh/
- Twitter/X: https://x.com/ain3sh
- Google Scholar: https://scholar.google.com/citations?user=j87_K9EAAAAJ
- Blog: https://ain3sh.com/blog
- Resume: https://ain3sh.com/resume

---

*This content is optimized for LLM consumption. For the full interactive experience, please visit https://ain3sh.com*