Building high-throughput backend services, deterministic execution kernels, and zero-copy data pipelines.
- Sub-microsecond Decision Kernel: Built integer-only deterministic routing kernels in Rust executing at ~115 ns per decision (8.6M decisions/sec) with zero heap allocation.
- Zero-Copy Arrow Pipelines: Engineered data validation and contract enforcement running directly on Apache Arrow C Streams (
Utf8View/BinaryView) with bounded memory. - Out-of-Core Spill Engine: Designed external K-Way merge-sort spill engines enforcing strict RAM budgets (e.g. 64 MB) over multi-million row datasets without OOM crashes.
- Cryptographic Audit Trails: Implemented BLAKE3 canonical dataset fingerprinting and Ed25519 digital evidence receipts for verifiable compliance (SOC 2, ISO 42001, EU AI Act).
- Formal Correctness & Concurrency Safety: Enforced
#![forbid(unsafe_code)], multi-platform CI matrices, and validated concurrency boundaries using Loom, Miri, and Property-Based Testing (proptest).
🦀 Calybris Core — Deterministic, Proof-Carrying Decision Kernel
A high-integrity decision primitive for LLM routing, financial guardrails, and budget-bounded automation.
- Evaluates constraints over frozen catalogs and emits verifiable proof bundles.
- Replay-auditable execution with byte-exact state transitions and a standalone
calybris-verifyauditor CLI. - Hash-chained WAL with HMAC signing and CAS budget accounting verified with Loom.
cargo add calybris-core🏹 ProofFrame — Arrow-Native Data Contract & Evidence Engine
A high-performance "Ruff-like" data validation kernel designed for high-throughput ETL and production pipelines.
- Compiles data contracts into typed Arrow kernels with zero-copy stream ingestion (Pandas, Polars, PyArrow).
- Memory-bounded out-of-core spill engine prevents Kubernetes pod OOM evictions on large datasets.
- Generates BLAKE3 dataset fingerprints and Ed25519-signed proof receipts for data lineage and compliance.
pip install proofframe==0.5.0| Project | Tech Stack | Business & Architectural Problem | Key Delivery & Output |
|---|---|---|---|
| ProofFrame | Rust, Apache Arrow, PyO3, BLAKE3, Ed25519 |
Production pipelines need sub-millisecond validation without OOM memory spikes. | Arrow-native engine, memory budgets, spill-to-disk exact state, signed receipts. |
| Calybris Core | Rust, PyO3, WAL, Loom, Miri, Ed25519 |
High-stakes automation requires deterministic decisions and tamper-evident audit trails. | 115 ns integer kernel, replay verification, single-writer WAL, CLI auditor. |
| Churn & Retention Report | Python, scikit-learn, SHAP, Pandera, Typst |
Raw churn probabilities fail to provide actionable operational retention strategies. | Calibrated risk scoring, XAI drivers, Pandera contracts, Typst executive PDF reports. |
| Scrape Quality Pipeline | Python, selectolax, asyncio, Pydantic v2, Pandera |
Web data ingestion breaks silently when selectors, schemas, or network rules shift. | Config-driven scrapers, polite rate limiting, Pydantic validation, Parquet/CSV export. |
| Price Monitor Pipeline | Python, Pandera, Typst, CLI |
E-commerce teams require automated competitor price monitoring and alert queues. | Watchlist config, threshold alert tables, run manifests, automated PDF summaries. |
I partner with engineering teams, startups, and enterprises for focused contract roles:
| Domain | Deliverables & Scope |
|---|---|
| Rust Systems Engineering | Performance-critical backends, Tokio/Axum services, deterministic state machines, memory profiling, and concurrency hardening. |
| Data Infrastructure & ETL | Arrow-native data quality pipelines, schema contracts, out-of-core processing, async scrapers, and automated reporting. |
| AI Cost & Spend Governance | Model gateways, token budget enforcement, fallback routing, policy simulation, and audit receipt architectures. |
| Applied ML & Decision Systems | Churn prediction pipelines, uplift modeling, SHAP explainability, calibration curves, and ranked operational queues. |
| Technical Audit & Refactoring | Concurrency verification (Loom/Miri), CI/CD pipeline repair, memory leak diagnosis, and production readiness audits. |
Languages : Rust (1.85+), Python (3.10–3.13), TypeScript, SQL
Backend : Tokio, Axum, FastAPI, Litestar, Pydantic v2, REST, OpenAPI, gRPC
Data & Storage: Apache Arrow, Polars, DuckDB, Parquet, PostgreSQL, SQLite, Redis
Testing & QA : Loom, Miri, Property-based testing (proptest), Fuzzing, pytest, Ruff
Delivery & Ops: Linux, Docker, Helm, Kubernetes, GitHub Actions, CI/CD, Typst
If you are dealing with an unstable data pipeline, expensive AI inference costs, a memory-intensive backend, or need auditable systems engineering:
- Business Problem & Context
- Current Tech Stack & Data Sources
- Expected Deliverable & Timeline
Send an inquiry to emirhuseyininci@gmail.com or connect on LinkedIn.