Shirin Shahabi
I am a Machine learning Researcher at Inference Labs, working on distributed verified inference. I have been working on Reinforcement Learning applications for the past 4 years, but currently I am more focused on Inference verification and Post training!
- Verifiable Inference and Privacy Preserving ML
- Compression, Optimization and a -bit- Inference Quantization
- Reinforcement learning and Finetuning
- Distributed inference
I am always looking to connect with people and discuss potential collaborations(the root how knowledge form and transfer) , Let's have a chat!
Skills
Featured Projects
Inference Labs - Verifiable AI & LLMs
- Distributed Proving Framework - Contributed as a core researcher to DSperse - open-source framework for distributed, verifiable inference. [paper] [sourcecode]
- JSTProve Integration - Contributed to JSTProve [paper] development and integration with DSperse, a specialized zkML toolkit built on Polyhedra Network's Expander backend to enable AI developers and ML engineers to generate and verify proofs of AI inference.
- AI Browser Agent Security - Devised and implemented Prompt Injection detection and zk-proxy systems for AI browser agents, leveraging DSperse masking for secure LLM source verification; instead of Prompt Guardlines.
- Commercial Aviation Vision Model - Developed distributed, verified inference for an external in-production, commercial Vision aviation model using DSperse and JSTProve preserving Inference fidelity, served as auditing and compliance tool.
- Quantization and Post-training - Contributed to an open-source ZK toolkit for quantization and post-training methods.
- Verifiable AI Agent - Developed the first verifiable Doom agent, reducing overhead from 1000x to 8x.
- Modular Proving Framework - Designed a scalable proof-of-inference framework leveraging agentic testing and probabilistic proofs, with no computational overhead, using a JIT design pattern and caching system.
- DPO-Based Ranking - Implemented ranking profiles using Reinforcement Learning benchmarking DPO, Group Relative Policy Optimization (GRPO) and Reinforcement Learning Teacher (RLT).
Nobitex - Data Science & Finance
- Customer Clustering at Scale - Developed scalable, ensemble tree-based algorithm leveraging Spark for processing 4M+ customer transactions, driving MAU growth to 8.5M within one year
- Chain Fraud Detection Dashboard - Developed the Fraud pipeline discovery and classification in OTLP wallet transactions based on blockchain transaction monitoring
- Marketing Data Pipeline - Led cross-team collaboration to design an end-to-end Marketing Data Modeling ETL Pipeline with ERD system design
SnappMarket - Retail Tech & Optimization
- Autonomous Inventory System - Developed scalable inventory reordering system for 15 hypermarkets with +8,000 SKUs
- Automated Shopping Experience - Contributed to a $100M funded project for a fully automated shopping experience (Low-Cost Amazon Go), implementing indoor location tracking in collaboration with Rocket Internet
- Operational Analytics - Established real-time operational KPIs with automated reporting systems, ensuring data accuracy and driving strategic decisions
Education
M.Sc. Computer Science
- Thesis: Stochastic Weight Optimization leveraging hierarchical reinforcement learning in Multi-objective network.
- Member of DFIC Quant Group - Quantitative Finance Department
- Participated in RLDL Vector Summer School at Vector Institute
MBA, Finance
Direct admission with academic ranking. Merit-based admission by the Office of Exceptional Talents.
news
| Oct 28, 2025 | JSTprove Paper is Live! |
|---|---|
| Oct 01, 2025 | Accepted to RL Residency at Prime Intellect. |
| Aug 09, 2025 | First DSperse Paper and framwork is out! |
| May 01, 2025 | Promoted to AI Researcher at Inference Labs! Starting May 2025, I will be leading research initiatives in distributed verifiable inference and zero-knowledge machine learning. |
| Mar 19, 2025 | Our paper titled “Enhanced Pareto Optimality with Reinforcement Learning Approach” has been accepted for presentation at the CORS 2025 Conference. |
latest posts
| Oct 28, 2025 | GKR for LLM Inference Proving — A New Era for Verifiable AI |
|---|---|
| Jul 10, 2024 | Can We Verify Every Action of an AI Agent? |