US
for US
- Location
- Charlotte, NC 28273
- Phone
- 2168487236
- Address
- Charlotte, 28273, United States
- Document
- Sahil Resume 2026 PM.pdf
One per right to work. The default is what a posting falls back to when its country matches none of them.
for US
for IN
The resume editor is off on this deployment. To turn it on, set JOBBY_OWNER_PASSWORD (at least 12 characters) and JOBBY_SESSION_SECRET (at least 32 random characters) in the Vercel project's environment variables and redeploy.
Every question answered once, kept for the next form. 200 stored
Story: earning trust at program scale (75-customer migration)
Competencies: leadership, stakeholder management, integrity, delivering results. Owned migration of 75 enterprise customers from competing FBO platforms. Found feature gaps that would break live aviation operations for some customers if moved early. Chose candor: told leadership and customers the specific gaps, ran a feature-parity program, re-sequenced migrations, communicated slips before being asked. All 75 migrated with zero downtime and zero churn on a forced platform change; 104% YoY ARR growth two years running.
Story: 10x-ing technical delivery at CAMP Systems
Competencies: delivery transformation, layered bets, operating cadence. Over ~18 months at CAMP: on arrival the team shipped roughly once a month, QA coverage was 62%, and the platform was a monolith creaking as the customer base scaled. Three layered bets. (1) LaunchDarkly: he owned the full rollout, not just licensing but integration patterns, change management and team enablement, taking monthly deploys to 4x release frequency within months. (2) AI tooling on the right bottleneck: the Invoicing module carried a 7-month estimate; he mapped where the complexity actually lived, used Cursor and AI workflows to accelerate scaffolding and boilerplate, and it shipped in 2 months. (3) QA automation: coverage 62% to 81% WHILE migrating monolith to microservices, which is the hard direction since coverage usually drops during migrations. Underneath all of it, a zero-downtime migration of 75 enterprise customers with dependency mapping, sequenced rollouts, rollback procedures and real-time stakeholder communication. No production incidents.
Story: delivering results under an impossible constraint (ETL automation)
Competencies: delivering results, process automation, prioritization. At CAMP, onboarding one enterprise customer onto TFBO Next-Gen took 2+ weeks of manual ETL. Owned a 75-customer migration program; sequentially that is ~3 years, mathematically impossible. Drove automation of the data-injection ETL instead of asking for headcount, taking capacity from 2+ weeks per customer to 3 customers per day. Made the program feasible and fed the 104% YoY ARR result.
Story: initiative and frugality (feature-flag governance web-app)
Competencies: initiative, frugality, technical judgment. LaunchDarkly's dashboard was flag-centric while our workflow was release-centric, and seats were expensive. Used Cursor for end-to-end development of a LaunchDarkly-API JavaScript web-app for tracking and controlling flag releases; consolidated seats and cut team OPEX 7%. Nobody asked for it; claimed the problem because I could see it.
Story: driving adoption and influence without authority (LaunchDarkly rollout)
Competencies: change management, influence without authority, learning from feedback. Drove org-wide adoption of LaunchDarkly, decoupling deployment from release; feature launch frequency up 400%. Honest failure beat inside it: early rollout was built around my own mental model and adoption stalled; feedback was that I shipped my tool, not their tool. Watched real users work, rebuilt around their friction, adoption followed. Standing lesson: never ship an internal tool without watching a real user attempt the real task first.
Story: root-cause conviction against consensus (monolith to microservices)
Competencies: technical depth, data-driven conviction, backbone. Enterprise customers hit timeouts on the legacy monolith; consensus said more infrastructure. Traced request-level timing and found synchronous retrieval paths and no caching, an architecture problem not capacity. Led transition to domain-driven microservices (Java, Spring Boot, TypeScript, Angular) with reusable REST APIs, caching, lazy loading targeted at hot paths. 68% reduction in data-retrieval latency, timeouts eliminated, cheaper than the infrastructure path.
Story: customer obsession with executive stakes (Million Air CEO digest)
Competencies: customer obsession, ownership, data integrity. Million Air, second-largest account, 25 airport locations: CEO received 25 separate weekly revenue emails from his FBO managers. Engineering had no bandwidth, so volunteered and built a dynamic PowerBI digest (cumulative revenue, per-location drill-down) auto-delivered every Sunday, verifying every figure against raw Snowflake data before it reached a CEO inbox. 25 messy emails became 1 verified digest at zero engineering cost; escalation became a strengthened relationship.
Story: 30-second automation color (portfolio chatbot)
Competencies: automation reflex, levity. People kept emailing or scheduling calls to ask basic questions about my background, so I built a chatbot into sahilsapra.com that answers in my voice instantly; recruiters get answers at 11pm. Use as a light second example, never lead with it.
Story: applied AI with real engineering depth (RAG test-case generator)
Competencies: AI product depth, quality engineering, innovation. Built an AI test-case generator grounded via RAG on ~2,000 of our own ticket-to-test conversions (Upstash vector index, dense embeddings) so outputs followed house conventions and improved with each added example, no retraining. Part of driving test coverage from 62% to 81% while holding enterprise release reliability. Can explain RAG, embeddings, and prompt iteration to both stakeholders and engineers.
Story: self-driven rigor and frugality (Skeptic backtesting engine)
Competencies: frugality, intellectual honesty, self-direction. Manual options backtesting on TradingView took hours per strategy and covered ~2 weeks of data, statistical noise. Built Skeptic, an adversarial backtesting engine designed to break strategies rather than confirm them, hunting curve-fit failure modes; own data collection and hosting under $20/month versus costly subscriptions. Guardrail when telling it: engineering and discipline, no options mechanics or P&L talk.
Story: founder-scale ownership and ambiguity (SpecHawk)
Competencies: think big, ownership, zero-to-one, fundraising. Founded SpecHawk, an AI-powered deploy-aware no-code agentic QA platform: maps staging environments, creates and executes test specs, validates Jira tickets with one click, screenshot evidence. Scoped MVP to the smallest slice producing a trustworthy test run on a partner's real app. Secured 4 enterprise design partners (their phrase: dream QA engineer) and raised $100K pre-seed via SAFE at $1M cap. Ran weekly design-partner loops measuring real scheduled usage, not demo enthusiasm.
Story: SpecHawk technical architecture and what broke
Competencies: technical depth, systems design, honest retrospection. SpecHawk architecture: TypeScript monorepo (Turborepo/pnpm), Next.js frontend, NestJS API, PostgreSQL with Prisma. Execution layer is Playwright workers inside BullMQ/Redis job queues so test runs are fully async. Flow: GitHub webhooks trigger staging discovery, then AI test generation, then queued Playwright jobs, then screenshot evidence back to the user. AWS throughout: ECS, Lambda, S3, CloudWatch. Deploy-aware is the core idea: it detects what changed in a deployment and tests that, rather than running the whole suite. WHAT BROKE, three things he says plainly: staging discovery was far more fragile than expected because companies do not have consistent staging URL patterns; early AI test generation was shallow, producing syntactically correct tests that tested the wrong things; the job queue backed up when several partners deployed at once because priority and background runs shared one queue. WHAT HE WOULD DO DIFFERENTLY: build the AI test-quality evaluation framework before anything else, since every V1 prompting decision was made on feel rather than data and they paid for it in rework; and treat staging discovery as a first-class problem from day one instead of deprioritising it for not being glamorous.
Answers are stored on Your Device and mirrored to the board behind this password so you can read them here, including compensation and demographics: this is your board and your data. That changes when a second account exists, and the per-user encryption in the multi-tenant plan is what changes it.