.png&w=256&q=75)
Thair Hasan@Thairhasan_ai2026
5
Events attended
3
Submissions made
10+ years of experience
About me
Software Engineer with over 10 years of experience in designing, developing, and maintaining high-quality software solutions. Proficient in backend development, with a strong focus on Java. Skilled in collaborating with cross-functional teams, identifying and resolving software-related issues, and staying updated with the latest industry trends. Known for boosting project completion, improving code quality, and delivering best user experiences.
🤓 Latest Submissions
AEGIS is a self-evolving WAF
Static WAFs only block attacks someone already wrote a signature for. AEGIS closes that gap autonomously. Traffic our rules can't explain gets embedded and clustered on an AMD Instinct MI300X, then sent to an LLM (also GPU-hosted) that proposes a regex and rationale. A deterministic Rust engine tests the candidate against real attack and benign corpora nothing ships on an LLM's opinion alone. Validated rules hot-reload live, no restart, with a human-review mode available. Underneath sits a production-grade Rust/Pingora core: OWASP-aligned rules, DDoS/rate limiting, sub-30ms decisions, and a dashboard with live gauge telemetry per AI signal. Three AMD GPU workloads power it, LLM inference, continuous model training, large-batch embedding benchmarked against CPU, not just claimed.
13 Jul 2026
PlamyraAI trade agent for surge hackathon
Overview Fully autonomous agent with a 6‑agent AI council analysing live Binance data (order book, trades, on‑chain, sentiment) to produce BUY/SELL/HOLD. Submits EIP‑712 signed TradeIntents to RiskRouter (Sepolia), posts validation checkpoints, enforces risk limits, and auto‑closes positions on TP/SL/timeout via Redis‑persisted open trades. Architecture Java (Spring Boot): Orchestration, Binance WebSocket (depth10,bookTicker,aggTrade), Redis, Web3j, EIP‑712 signing. Python (FastAPI): 6‑agent council (Technical, Macro, On‑Chain, Risk, Devil’s Advocate, Microstructure) using Llama3 (Ollama), regime detection, Kelly sizing. Redis: Open position store (survives restarts, enables TP/SL recovery). Blockchain: Sepolia – AgentRegistry, RiskRouter (0xd6A69525…), ValidationRegistry, HackathonVault. Trading Flow Java builds MarketState (price, RSI, ATR, MACD, Bollinger, cumulative delta, fear/greed). Python council returns verdicts → judge synthesises decision. Risk engine applies confidence floors, spread/R‑R guards, Bayesian scaling. Java signs TradeIntent, submits to RiskRouter.submitTradeIntent() (max $500/trade, 10/hour, 5% drawdown). Posts signed checkpoint to ValidationRegistry with score (0‑100). Open trade saved to Redis, monitored every 500ms. On TP/SL/timeout, PositionCloserService submits SELL intent on‑chain. Key Features Regime‑aware prompts (Bull, Bear, Ranging, Crisis, Accumulation, Distribution). Real‑time microstructure (depth, cumulative delta, VWAP, spread). Dynamic TP/SL (ATR × regime multiplier). Validator‑ready: signed checkpoints + local checkpoints.jsonl audit trail. Tech Stack Java 17, Spring Boot, Web3j, Lettuce (Redis) | Python 3.11, FastAPI, Ollama, ChromaDB | Binance WebSocket | Sepolia testnet | Gradle / pip. Repos Java: https://github.com/thassan1977/surge-hackathon Python: https://github.com/thassan1977/surge-ai-agent Agent ID: 36 Operator wallet: 0xED1e806796A98725D5B3A07478440977dBE34812
12 Apr 2026
