About
Researcher, Engineer & Long-Term Systems Builder
Before focusing on AI systems research, I spent more than fifteen years building production software, leading technical work, and taking products from initial concept through deployment and iteration. I later completed a B.Eng. in Computer Systems Engineering and an MASc focused on compound LLM agents, prompt-only adaptation, and reliable sequential decision-making.
That path shapes how I approach research. I care not only about whether a method works in one demonstration, but also about the infrastructure around it: what information the model receives, how experiments are reproduced, how failures are measured, how cost is accounted for, and how a result can survive contact with a working system.
- LLM agents
- Post-training: SFT · LoRA · RL (PPO, DPO, GRPO)
- Multi-agent RL
- Agent evaluation & reliability
- Mechanistic interpretability
- Transformers
- PyTorch · JAX
01
Current research focus
Reliable agent systems
How compound LLM agents should be structured: what they see, how decisions are decomposed, and where deliberation helps or hurts.
Evaluation
Time-to-event methods, failure taxonomies, and cost accounting that reveal when and why agents stop working.
Adaptation
Prompt-only and inference-time methods that let agents improve from experience without weight updates.
Interpretability
Sparse-autoencoder and causal analysis of whether reasoning representations are shared across languages.
02
Professional trajectory
Graduate AI Researcher · Carleton University
Realistic and generalizable training of autonomous cyber agents. Designed and implemented compound LLM-agent systems for decision-making in complex, stochastic, partially observable environments. Built the experiment infrastructure.
- Two first-author ACM CAIS 2026 papers with all ACM reproducibility artifact badges.
- Invited research poster presentation of the ACM CAIS work at AI Engineer World’s Fair 2026, San Francisco.
- First-author ICML 2026 Mechanistic Interpretability Workshop and ICLR 2026 workshop papers.
Founder & Technical Director (part-time) · Appalect
Architecture and hands-on development across web, mobile, backend, automation, and integration systems, from concept and prototyping through deployment and iteration.
- Applied AI/LLM tooling and research infrastructure: multi-provider inference, experiment orchestration, configuration, retries, logging, and evaluation workflows.
Teaching Assistant · Carleton University
Intro to Machine Learning (designed assignments and the course project, weekly hands-on AI/ML/DL tutorials for 120+ students), Computer Systems Lab, Real-Time Concurrent Systems, and Intro to Python.
Founder & CTO · IBCICO Development
Founded and led a full-cycle software development agency, managing an engineering team through the entire software development life cycle building financial data systems, e-commerce, and ticket-booking platforms.
03
Education
- 2024 – 2026
Master of Applied Science (MASc)
Electrical and Computer Engineering
Carleton University · Ottawa, Canada
- CGPA 11.6/12.0 (GPA 4.0/4.0)
- Queen Elizabeth II Graduate Scholarship in Science and Technology
ThesisDesign and Gradient-Free Adaptation of Compound LLM Agents in a Partially Observable Adversarial Networked Environment (opens in a new tab) · Nominated for the Senate Medal
- 2019 – 2024
Bachelor of Engineering (B.Eng.)
Computer Systems Engineering · minors in Mathematics and Physics
Carleton University · Ottawa, Canada
- High Distinction, CGPA 10.78/12.0
- Dean’s List and Scholarship: 2020, 2021, 2022
- 2000 – 2005
Specialist Diploma (five-year)
Linguistics and Intercultural Communication
Moscow State Regional University · Moscow
- Cumulative GPA 4.7/5
04
Awards and achievements
Queen Elizabeth II Graduate Scholarship in Science and Technology
Awarded to 19 graduate STEM students at Carleton University, 2025.
MASc thesis nominated for the Senate Medal
Carleton University, 2026. CGPA 11.6/12.0 (GPA 4.0/4.0).
B.Eng. with High Distinction
Computer Systems Engineering, Carleton University, 2024. CGPA 10.78/12.0.
Double minor in Mathematics and Physics
Completed alongside the engineering degree.
Dean’s List and Scholarship
2020, 2021, and 2022.
ACM reproducibility artifact badges
All badges awarded for both ACM CAIS 2026 papers.
Invited research poster presentation, AI Engineer World’s Fair 2026
Research Papers track, San Francisco, invited after the ACM CAIS papers.
VolunteeringReviewer for NeurIPS 2026, GLOBECOM 2026, ICLR 2026 LLA, and COLM 2026 LLA (February 2026–present).
05
Invited talks and presentations
- June 29 – July 3, 2026
Invited research poster presentation
Compound LLM agent design and FORGE: prompt-only memory evolution
AI Engineer World’s Fair 2026, Research Papers track · San Francisco
Invited after the two ACM CAIS 2026 papers.
Post (opens in a new tab) - May 2024
Short talk
Video-Based Clinical Intervention Detection and Classification with Summary Report Generation Using Transformer Models
Montreal Child CDSS Congress (M3C) · Montreal
06
Working style
- Empirical
- Claims are tied to controlled experiments, ablations, and replication across model families.
- Hands-on
- I write the harness, the connectors, the runner, and the analysis, not only the paper.
- Systems-oriented
- A method counts when the infrastructure around it (config, logging, retries, reproduction) also works.
- Cost-aware
- Tokens, calls, and wall-clock are first-class metrics, reported alongside accuracy and return.
- Across research and production
- Fifteen years of shipping software shape how I scope, build, and hand off research systems.
- Direct about uncertainty
- Every project page on this site has a limitations section. Workshop papers are labelled as workshop papers.
Contact
Building reliable AI systems requires both research and engineering.
I’m interested in research engineering, applied research, agent infrastructure, evaluation, reliability, interpretability, and research-to-production work.