Systems & Research Engineer

**E3 Group** Systems & Research Engineer; $220K–$300K TC (cash + equity); Remote (SF office available)

**About Company** E3 Group (the name stands for Explore, Experiment, Expand) is an applied AI lab building systems for the freight and supply chain industry — fraud detection, high-volume voice and document processing, and carrier/customer communications. Co-founded by Caleb Sirak (MIT; background in data-center chip design and deployment, previously scaled a business to ~$24B in annual volume) and Aqil Naeem (Stanford; working in supply-chain distribution since age 13), E3 raised a $4.5M seed round in January 2026.

Its first product, Shield, is a driver-fraud and identity-verification platform now screening ~5,000 drivers/day (300K+ ingested industry-wide). The team has since shipped Frontline, a high-volume voice-agent product running at roughly 10x lower cost than competitors with better transfer rates, plus workflow-organization and document-ingestion products — with a roadmap to expand across freight AI workflows over the next 5-10 years.

The team is flat: everyone holds the title "member of technical staff," intern to C-suite, and the culture is built around heavy day-to-day use of coding agents, with engineers expected to operate with founder-level ownership across the stack.

**About the Role** The Systems/Research counterpart to the Applied seat — internal-facing, focused on the infrastructure and research that Applied engineers deploy to customers. A recent example: benchmarking speech-to-text providers and building a hybrid open-source/production system that beat vendor options on both cost and quality.

**Responsibilities** - Own performance and scale for E3's core systems (Shield, Frontline voice infra, document ingestion) — profiling, optimizing, and re-architecting as volume grows. - Run structured research: form a hypothesis, design a benchmark, gather the data, and make a defensible call — e.g., evaluating and combining open-source vs. closed-source model providers. - Reason about distributed-systems tradeoffs at scale (concurrency, latency, cost) wherever LLMs or other models are being served. - Communicate findings and system constraints clearly to the Applied team so they can plan around them. - Use coding agents as the default way of working: spend your judgment on what to build and verify rather than on manual execution.

**Required Skills** A. Performance engineering on a model-system mechanism: inference, serving, evaluation, retrieval, speech, or agent infrastructure. Distributed-systems work with no model in the loop does not qualify. B. Rigorous, scientific-method reasoning — can explain the hypothesis behind a technical decision, what alternatives were considered, and how bias in the data/benchmark was addressed. C. The research itself must be on an AI system: model serving, benchmarking, or evaluation at scale, with the concurrency, latency and cost tradeoffs that come with it. D. Heavy, fluent daily use of coding agents — assessed directly in the interview process.

**Bonus Skills** - AI/ML research background specifically. - Physics or other quantitative-science background.

**Logistical Info** - Location: E3 is HQ'd in San Francisco; in-office is welcome but not required — open to any geography for the right talent. - Compensation: $220K–$300K total comp (cash + equity) for a senior hire; flexible upward for a candidate who can clearly justify it. - Number of openings: Ongoing pipeline (opportunistic, high-bar hiring — no fixed headcount target). - **Other:** Fully async-friendly — no fixed working hours or overlap requirement (E3 runs 24/7 team coverage globally). Strong written/async communication expected: post updates to a shared channel as soon as things happen.

**Ideal Background** - Builders with visible side projects, even small or incomplete ones — E3 weights independent curiosity and self-directed exploration heavily. - Candidates who can precisely explain what they built and why, and name the real bottleneck they solved — not just the scale of what they touched. - Physics or other quantitative-science backgrounds moving into software are a positive pattern E3 has seen work well. - Comfortable presenting a benchmark-driven decision end to end — from hypothesis to data to recommendation. - Comp note: Caleb's guidance is ~$240K–$300K TC for a senior hire; anything meaningfully above $300K needs an exceptional, clearly-communicated case — treat $300K as the practical ceiling, not the target. - Process note: submissions go via Slack to Naomi Birman (E3's recruiting ops lead, new to the role, still building out Notion/Slack automations) — expect some early process friction while that beds in.

**Green Flags** - Multiple varied, self-taught technical projects (range and depth over one polished credential). - A "founding engineer" or leadership title at a recognizable company — worth the conversation even if the resume itself is thin on detail. - At least one bullet where you can trace a decision from problem choice, through verification, to a real outcome. Metric density is not the test. - Approaches problems from data, not belief — can show how they validated a call rather than just asserting it.

**Red Flags** - "Jargon-maxing" — inflated resume language; can't explain what they actually did or why when asked directly. - Numbers with no baseline, no method, and no way to check them. Internally inconsistent resumes. Concurrent affiliations that cannot all be real. - No relevant AI/agent experience at all. - Comp expectations meaningfully above ~$300K TC without an exceptional case for it. - Operates on belief/intuition rather than data; can't explain how they'd validate a decision. - Big-company IC background with no founder, founding-engineer, or 0-to-1 stretch. Presumptive no unless the submission supplies context that overturns it.

**Interview Process** 1. Resume/intro sync (~15 min). 2. Behavioral conversation and interview-process walkthrough (~15 min). 3. Coding round (~1 hr), including a hard requirement to demonstrate fluent coding-agent use. 4. System design round (~1 hr). 5. Final round with co-founder Caleb Sirak.