I design and build production-grade agentic AI systems and study how large language models fail in extended interactions.
I treat LLMs not as magic boxes but as stochastic components that require deterministic guardrails, typed state management, and structured orchestration. My work sits at the intersection of production systems engineering and applied AI research — building the systems, then studying their failure modes.
🟢 Status: Graduating August 2026 — open to AI engineering roles and Masters applications (Winter 2027/28 intake)
| 🏗️ The Builder | 🔬 The Researcher |
|---|---|
| Production multi-agent systems with typed state, parallel execution, quality gates, and behavioral safeguards | Documented behavioral research on production LLM workflows, with research interests in context anchoring and long-horizon drift |
| LangGraph · Google ADK · FastAPI · Flutter · SSE Streaming | Context Engineering · Red Teaming · Token Sampling · A/B Testing |
| Project | Description | Stack |
|---|---|---|
| Rahzaan — AI Career Guidance (Private FYP · Live Demo) | Production-deployed career counselling platform. 26-node parallel multi-agent pipeline cutting first-response latency from 20–25 seconds to 3.7 seconds, 265-question psychometric assessment stack, live SSE streaming, 29-route Flutter frontend. | LangGraph FastAPI Flutter Supabase |
| Autonomous Research Concierge | Production multi-agent research pipeline. An intent synthesiser compacts chat history into a research directive before three search agents fan out concurrently. Per-call session isolation, deterministic security guardrail, draft-critique-validate loop. Deployed on Railway. | Google ADK Gemini API Docker Python |
| Applied LLM Workflow Research | Independent behavioral research on Gemini 2.5 Pro. Documents three original constructs — Optimal Answer Trap, Final Synthesis Bias, Conversational Gravity — with controlled case studies and quantitative findings. | Gemini API Streamlit Python |
| Crypto Accumulation Scanner | Scans 1,200+ cryptocurrencies daily across CoinMarketCap and MEXC Futures. Classifies price action into 7 accumulation signatures and applies a log-log regression Z-score volatility model derived from 2020–2024 market data, normalised by market-cap tier. | Google Apps Script JavaScript CoinMarketCap API |
| WordPy Solver | Autonomous Wordle solver using pixel colour sampling to read game state without OCR. Constraint-based filtering with a greedy information-gain heuristic. 85%+ solve rate within six guesses across a 5,922-word list, covered by a 36-test suite. | Python Pillow |
Agentic AI & LLM
Languages
Backend & Infrastructure
Frontend
Tools
AI Engineering Practices Context Engineering · Prompt Engineering (Red Teaming, Chain of Density, Tree of Thoughts, Token Sampling)
Stats images generated by jstrieb/github-stats.
- Rahzaan — finalised as a final year project, now continuing as a student-facing product with expanding university dataset coverage
- Independent research interests in LLM behavioral failure modes: context anchoring and long-horizon drift
- Preparing Masters applications for AI and intelligent systems programs — Winter 2027/28 intake
