Tirth Joshi
Artificial Engineer and International Tutor
Tirth Joshi
M.S. Artificial Intelligence, Yeshiva University (GPA: 4.0) · New York City
Research Assistant × 3 · Published at IEEE · Under review at ACM FAccT & ECCV
I build AI systems that have to work under real constraints — adversarial inputs, incomplete information, and non-stationary environments. My work sits at the intersection of LLMs, reinforcement learning, graph representations, and privacy theory.
Current Research (Dec 2025 – Present)
I hold three concurrent research assistantships at Yeshiva University:
Prof. Shucheng Yu — LLM Security
Building a synthetic dataset generator for adversarial smart contract samples targeting underrepresented vulnerability classes, including zero-day patterns. Developing an LLM-based detection framework with RL-style feedback loops that iteratively refine detection policies from sparse execution signals.
Prof. Honggang Wang — Multi-Agent RL for IoT
Designing an RL-based multi-agent coordination system for heterogeneous IoT environments — a humanoid robot and a remote-controlled vehicle jointly optimizing under partial observability, sensor fusion, and real-time constraints. Addressing inter-agent communication and non-stationarity across vision, audio, and motion modalities.
Prof. Aaron Ross — Compositional Privacy Risk
Developing metrics that quantify re-identification potential from cumulative data releases. Formalizing how anonymity erodes as datasets and models accumulate combined signals over time — even without explicit identifiers. Work submitted to ACM FAccT.
Publications
| Hierarchical Graph Representation for Multi-Chain Blockchain Routing | IEEE ICNC 2026 — Invited Paper ✅ |
| Measuring the Statistical Erosion of Anonymity: RPI | ACM FAccT — Under Review |
| Formalizing Compositional Privacy Risks in Non-Face Re-Identification | ECCV — Under Review |
| Soft-Landing Liquidations for Overcollateralized Lending | DeDeFi Workshop — Under Review |
| Do Phonetic Patterns Predict Grammatical Structure? | YU CSE Day & DuckAI 2025 ✅ |
→ Full research details and abstracts
Experience
Research Assistant — Yeshiva University · Dec 2025 – Present
Three concurrent faculty projects (see above).
AI Engineer — DZap, Bangalore · Sep 2024 – Dec 2024
Designed an inter-chain path-finding framework formulating cross-chain routing as sequential decision-making over dynamic gas costs and liquidity constraints — reducing gas costs by 27%. Built LLM-integrated blockchain agents enabling natural-language DeFi commands via multi-agent planning, cutting manual input time ~40%.
Machine Learning Intern — ComputeLib, Delhi · Jun 2023 – May 2024
Developed gRPC/REST microservices, improving API latency from 2.1s to 0.6s (~3.5× speedup). Deployed Hugging Face + LangChain LLM backend serving ~25k requests/day. Containerized modules with Docker, improving uptime by 43%.
Selected Highlights
- 🏆 Best Research Award — YU CSE Day 2025
- 📄 Featured in YU News for IEEE ICNC 2026 paper
- 🚗 Top 0.80% globally — AWS DeepRacer Student Qualifier (19 / 2,362)
- 🎓 AWS AI & ML Scholarship, L&T TS Scholarship (IIT Madras)
- 🌍 Global Guide, One To World — 150+ international students mentored
Teaching & Service
- Global Guide, One To World — mentored 150+ international students on academics and life in the US
- Section Leader, Code in Place — Stanford’s global introductory coding program
- Mentor, DevFest Hackathon — Columbia University
- Volunteer — Jersey City Marathon
Skills
Languages: Python · C/C++ · TypeScript · Java · Kotlin · SQL
ML / RL: PyTorch · TensorFlow · Scikit-learn · OpenAI Gym · Stable-Baselines3 · PPO / DQN / SAC
LLMs & Retrieval: LangChain · FAISS · ChromaDB · Pinecone · Qdrant · Hugging Face
Infrastructure: Docker · FastAPI · Flask · Django · Express.js · JWT
Databases & Graphs: MySQL · Firebase · Neo4j · The Graph
Writing
I write about ML, quantum computing, and systems on Medium. Selected pieces:
- Traveling Salesman Problem Using Quantum Computing — The Quantastic Journal
- Quantum Machine Learning for ML Engineers
- Integration of Quantum Computing with Classical Data Systems
I care about depth over breadth, systems over demos, and clarity over hype.
Favorite equation: the Riemann Hypothesis.
tjoshi1@mail.yu.edu · Jersey City, NJ