Dioptra (an Icertis Company) — Senior Software Engineer

• "Really an AI and a backend engineer." Python required; AI means agentic experience specifically.
• Owns the security requirements the company takes on: the in-progress SOC 2 audit (implementations, controls, infra), customer BYOK requests (customers holding their own encryption keys), and securing AI-agent infrastructure (e.g., the infra patterns and MCP surfaces that agents run on).
• A fair amount of infrastructure work alongside backend — AWS or Azure, scaling, and some DevOps.
• Intake framing was backend-only: "I don't expect them to be doing front end or UI." (Conflicts with the role page, which lists production JS/TS and end-to-end agentic incl. a front-end component as Required — see flag.)

• →5+ years building and shipping production backend systems. [Required]

• Shipped production systems at a startup or high-ownership environment. [Must have]
• Designed or implemented security-sensitive systems (encryption, key management, auth). [Required]
• Built or deployed AI/ML or LLM-powered agents in production. [Required]
• Must have worked in a small startup (under 100 people) while there — as founder, founding engineer, or early hire. [Required]

• No specific degree requirement stated in the role brief. (Intake, for the design role, noted non-CS/creative backgrounds as a plus; for this role, production engineering ability is what matters.)

• Strong Python — production-level code for ML and agentic infrastructure. [Must have]
• Cloud infrastructure experience (AWS or Azure) including DevOps. [Required]
• Used JavaScript or TypeScript in a production environment. [Required — note intake's backend-only framing]
• Building agentic systems end-to-end; understands agentic harness / MCP and orchestration well enough to do a systems design around them. [Required]

• High agency: self-directs, identifies and solves problems independently. [Must have]
• Comfortable translating ambiguous enterprise security requirements into concrete systems. [Required]

• Willing to work in-office in New York 5 days/week. [Must have]
• Now also accepting candidates in San Francisco, on-site 5 days/week. [Required]

• Primarily non-tech / non-product company background (e.g., banks, consulting). McKinsey is the rare exception.
• Slow-moving enterprise mindset (e.g., long tenures at Microsoft-like cultures). Meta-style move-fast is fine.
• Frequent job hopping (every six months).

• Salary: $180K–$280K (role page) / $180K–$270K (intake) — see comp flag
• Equity: Competitive equity
• On-site policy: 5 days/week on-site; NYC (Brooklyn office) or San Francisco. Intake said "at least 4 days," typical hours ~8:30am–7:30pm, explicitly not a 996 culture; the 5-day Must-have (updated) governs.
• Visa sponsorship: Open to relocating candidates; will transfer existing visas (e.g., H-1B, OPT transfers); will not sponsor brand-new visas.
• Employment type: Full-time
• Location: New York, NY (Brooklyn) or San Francisco, CA

Role page indicates 4 steps; intake described this 3-step flow. Confirm whether a stage was added. Doug is the interview-coordination point of contact.

No dedicated Ideal Companies section was present in the copied HTML. Signal from the strong-example profile and intake: backend/infra engineers with a security eye out of regulated or high-ownership tech environments (e.g., fintech), with a genuine startup/founding chapter and production AI/agentic work. Meta-style product companies read well; Microsoft and non-tech core businesses read poorly. Confirm with a dedicated Ideal Companies list from Paraform if one exists.

For reference only — do not source these specific profiles.

• Founding / startup experience: Only consider candidates with proven early-stage or founding engineering roles that owned end-to-end product features — not just big-tech backgrounds.
• Agentic product ownership: Must demonstrate hands-on experience building production-ready AI agent systems (MCP + agent orchestration) spanning both back-end and front-end.
• Relevant tech stack & location: Prioritize strong Python, TypeScript/React, and AWS, with full commitment to 5-day on-site in NYC or SF.

• "Not enough agentic AI experience" is the single most common kill reason — non-negotiable "until we get bigger."
• "Zero small-company / startup experience" and "no start-up experience" kill even strong big-tech engineers.
• "At least 18 months in a startup" wanted — very short startup stints may not clear the bar.
• "Not in NYC" cited pre-SF opening — location commitment matters (now NYC or SF, 5 days).
• Communication/fit rejections: "poor communication," "overly complicated thinking, may not do well in a startup," "not responsive."
• "Advising non-tangible founders seems pretty thin" — advisory/fractional roles don't substitute for hands-on founding/early-engineer work.
• One strong candidate rejected for wanting "one experience with a larger established company" — the large-company balance factor cuts both ways.

• Meta ML/ads engineer (Kril, via Para AI): rejected — no clear security work in recent roles despite strong ML.
• Microsoft identity-platform engineer: had genuine auth/access-control work, but rejected purely on the Microsoft "enterprisey and slow" concern.

• Startup experience is effectively a hard gate. Farah called it a "nice-to-have" for the backend role in the abstract, but the role page (updated) marks it Required and the rejection history kills big-tech-only profiles repeatedly. Score a missing genuine startup/founding chapter as a Must-have miss.
• Backend/infra architecture first, security second. Per Doug's course-correction, do not over-index on pure security-engineer titles; prioritize strong backend + infrastructure + data-systems architecture with a security eye.
• Agentic production experience is non-negotiable. Internal dev tooling for coding does not count — must be agentic systems built for a product, with MCP/orchestration understanding.
• Company pedigree filter: tech/product companies count; non-tech core (banks, consulting) generally doesn't (McKinsey excepted). Microsoft strongly disfavored; Meta-style favored.
• Backend-only vs full-stack conflict: intake said no front-end/UI; role page requires production JS/TS + end-to-end agentic incl. a front-end component. Treat the role page as governing but surface for HM confirmation.