Why this persona fits the audience
Works as an applied ML engineer on a 2-person AI pod at a logistics startup in London; she maintains a Python agent workflow that summarizes shipment exceptions and suggests actions, with traces in LangSmith and escalation checks kept in spreadsheets. The team is currently comparing several model backends in the same prototype.
- Age: 29
- Location: London, UK
- Occupation: Applied ML Engineer
- Income: £82k base + equity
- Education: MSc Data Science
- Company size: 15-20 employees (startup)
- Years in role: 4 years
“I don't trust the model until I've seen the trace log myself; otherwise, I'm just guessing.”
Erma Jerde, 29, is a quintessential modern applied machine learning engineer based in London, operating at the intersection of rigorous data science and gritty logistics reality. Holding an MSc in Data Science, she spent four years refining her craft before landing her current role in a lean, two-person AI pod at a fast-moving logistics startup. Unlike many peers who chase the latest hype cycles, Erma is driven by a deep-seated need for reliability and technical rigor. Her current mandate is to build a Python agent workflow capable of autonomously summarizing shipment exceptions and suggesting corrective actions. This project sits at the heart of the company's value proposition, yet it remains in a fragile state where the margin for error is non-existent. Her daily existence is defined by a constant tug-of-war between rapid iteration and the necessity of auditability. While the team prototypes new capabilities, Erma is responsible for the scaffolding that ensures those capabilities can be trusted. She maintains a sophisticated chain of thought traces using LangSmith, a tool she chose for its visibility features, and keeps a parallel set of escalation checks in Google Sheets—a testament to her belief that human oversight must remain a safety net even in automated systems. Currently, she is spearheading a comparative analysis of several model backends, weighing the trade-offs between inference speed, cost, and accuracy. This process is exhaustive; she runs dozens of test cases, manually verifying outputs against ground truth data. Financially, Erma earns £82,000 base plus a modest equity package, a compensation structure that reflects both the high stakes of her role and the precarious nature of the startup ecosystem. She views her income not just as a salary but as an investment in her own skill acquisition, frequently allocating funds to advanced workshops and hardware upgrades. In her personal life, she finds balance through the physical demands of rock climbing and the strategic complexity of board games, hobbies that mirror her professional approach to problem-solving. She lives in Hackney, sharing a space that doubles as a lab for her side projects, surrounded by the hum of servers and the scent of old books. Despite the chaos of the startup world, Erma remains calm, grounded by her conscientious nature and a clear vision of what constitutes a well-engineered system.
Background & daily life
A day in their life
Erma wakes at 6:15 AM, avoiding her phone until after a 45-minute run along the canal. Breakfast is quick—oatmeal and black coffee—while she skims ArXiv abstracts on transformer architectures. By 8:30 AM, she is in her home office, diving straight into code reviews for the previous day's agent logs. The morning is dedicated to deep work: refactoring the exception summarization logic and debugging a race condition in the LangSmith tracing pipeline. Lunch is a solitary affair, often a prepped meal while reading technical papers. Afternoon meetings are minimal, usually reserved for syncs with her co-founder on the backend comparison results. Late afternoon involves the tedious task of cross-referencing model outputs with manual spreadsheet checks. She ends her day by updating her personal knowledge base and planning the next day's sprint goals, ensuring no context is lost overnight.
Lifestyle
Erma lives in a converted warehouse flat in Hackney, characterized by exposed brick and a desk dominated by three monitors. Her routine is structured: early morning runs along the Regent's Canal to clear her head, followed by a strict work block. She prioritizes health through weekly bouldering sessions at a local gym, which she treats as non-negotiable mental maintenance. Financially, she is cautious but investing heavily in her career via conference attendance and specialized courses rather than luxury goods. Her social life revolves around low-key board game nights with friends or quiet dinners, preferring depth of conversation over large networking events. She manages her finances with automated savings and a strict budget for tech hardware.
Work life
Erma operates in a high-pressure, two-person pod where ambiguity is the only constant. Her days are split between debugging Python agents that handle shipment exceptions and manually validating outputs against spreadsheet logs. She spends significant time maintaining traceability in LangSmith to ensure auditability for stakeholders. Recently, she has been leading a critical evaluation of different model backends, running parallel prototypes to compare latency versus cost. The pressure comes from the need to deliver reliable automation without introducing hallucinations that could disrupt supply chains. Her workspace is cluttered with sticky notes and printed architecture diagrams, reflecting a mind constantly mapping out complex data flows.
Relationships
Erma is close with her mother, a retired teacher who calls every Sunday, providing a grounding counterbalance to her technical intensity. Her primary professional influence is her co-founder, a former operations manager whose pragmatic business sense complements Erma's technical focus. She has a tight-knit group of climbing partners in East London who know little about her job but provide essential stress relief. She is distant from extended family, preferring to curate her social circle carefully. Her decision-making is rarely influenced by peers, though she values the feedback of a small circle of trusted senior engineers.
How they decide
Erma employs a research-heavy, evidence-based decision style. She avoids impulsive choices, preferring to gather data points, run small-scale experiments, and consult documentation before committing to a path. She is skeptical of marketing claims and relies on benchmarks, community consensus, and peer-reviewed performance metrics. While she is open to innovation, she requires proof of stability and scalability before integrating new tools into her core workflows. She values transparency and will often choose a less feature-rich solution if it offers better reliability and explainability.
How they sound
Erma speaks with a precise, measured cadence, using British English idioms sparingly but effectively. Her vocabulary is technical but accessible, often defaulting to specific jargon like 'latency,' 'hallucination rate,' or 'trace' without explanation. She avoids hyperbole, rarely using words like 'amazing' or 'terrible' in favor of 'suboptimal' or 'inefficient.' She tends to pause before answering complex questions, organizing her thoughts internally first. She does not use slang unless joking with close friends and generally avoids emotional venting in professional settings, preferring to frame frustrations as solvable technical problems.
Personality traits
- meticulously organized
- quietly observant
- systematically skeptical
- deeply focused
- pragmatically optimistic
- methodical
- direct communicator
- curiosity-driven
- calm under pressure
- detail-oriented
Goals
- Deploy a fully autonomous exception handling system that reduces manual review by 80%
- Publish a case study on optimizing RAG pipelines for logistics data
- Achieve senior engineering status within the next 18 months
- Master a new edge-computing framework for on-device inference
- Build a sustainable work-life balance that prevents burnout during crunch times
Problems & frictions
- Maintaining data integrity when switching between multiple experimental model backends
- Lack of standardized documentation for legacy code written before the current pod structure
- Difficulty scaling the current prototype from a PoC to production-ready reliability
- Time scarcity caused by context-switching between coding, testing, and manual spreadsheet verification
- Uncertainty regarding the long-term viability of the startup's funding runway
- Managing the cognitive load of keeping up with rapidly evolving open-source tools
Values
- technical rigor
- learning fast
- reliability
- pragmatic efficiency
- intellectual honesty
Interests
- machine learning
- board games
- climbing
- analog synthesizers
- urban hiking
Media habits
- Hugging Face discussions and GitHub issues
- The Daily (podcast) for general news
- ArXiv.org daily digests
- Discord communities for specific LLM frameworks
- Bloomberg Tech newsletter