davidmiheev
Systems Engineer
I've worked on several research and commercial projects (from astronomical research software to quant research and recently infrastructure and AI engineering systems). I excited about current transition in software engineering because now we can tackle more complex problems than ever before. I open to wide range of opportunities - my main criterion is that you're solving complex and high-impact problem
Experience: 3 years
Yearly salary: $0
Hourly rate: $40
Nationality: 🌏 Remote
Residency: 🌏 Remote
Experience
Founding Engineer
dodil.io 2025 - 2026
- Designed and built dodil‑agent, a Rust agent platform: durable task‑DAG orchestrator on NATS JetStream KV and S3 with compare‑and‑swap task claiming, TTL‑leased worker membership and idempotent resume after crashes; gRPC and SSE APIs with event replay - Built sandboxed agent execution on NVIDIA OpenShell (Docker and Kubernetes compute drivers) behind a default‑deny egress proxy; per‑run token budgets enforced at an OpenAI‑compatible metering proxy - Co‑built Ignite, a serverless platform in Rust on which LLM executors and agent sandboxes run; wrote the candle‑based LLM executor (CUDA, quantization, HTTP/SSE serving); 35B MoE decode 2.5× faster (19.6 → 48.9 tok/s), cold start 65× faster (18.9 s → 0.29 s)
Quantitative Researcher & Developer
Freelance 2024 - 2025
- Built cross‑exchange arbitrage scanners in Rust: real‑time ingestion and processing of L2/L3 market data streams - Implemented and backtested GLFT market making with hftbacktest on recorded Binance order books (9 assets): calibrated fill intensity and volatility, modeled order queue, latency and inventory risk - Applied ML to signal discovery on recorded order books: built a markout‑based spread rule that beat baseline strategies on held‑out data across all strategy–asset pairs
Skills
ai
golang
python
rust
english