Redwood City, California, United States
I build and scale real-time AI platforms that win in the enterprise. Director of Engineering at Otter.ai, owning Otter Pilot, Data Platform, Reliability, Multi‑language ASR, and Real-time Speaker ID; 7+ patents. Known for turning research into revenue—shipping zero-to-one systems and scaling them to high adoption and strong margins. Otter.ai — Led the shift from mobile recording to a virtual meeting assistant platform. Delivered high–eight-figure ARR expansion and enterprise wins in finance, consulting, and media via sales-assisted real-time speaker identification, multi-language ASR, and purpose-built notetaker integrations. Achieved five 9s reliability for timely meeting joins; cut incidents by 45% and site outages by 67% through a reliability mandate and cross-functional pods. Negotiated multi-year vendor/cloud agreements and re-architected infrastructure for multimillion-dollar savings. Built a Webex workflow unlocking regulated fintech entry and driving meaningful seat expansion. Built and retained a high-performing org (~40 engineers/leaders) with exceptional retention. Meta/Facebook — Managed and tech-led a cross-functional team; shipped investigations tooling adopted across Facebook, Instagram, and WhatsApp. Reduced bad-actor recidivism by ~20% and cut time-to-insight from days to hours. Palo Alto Networks — Architected AutoFocus from inception to scale (zero data loss, DR across private DCs) at TB/day throughput. Helped grow it into a multi-million-dollar ARR product; earned multiple company-wide innovation awards. Cisco + early career — Led campus security features across Catalyst switch portfolios; earlier embedded/real-time roles underpin a reliability-first approach. 7+ patents in AI automation, security, and networking. Keywords: VP Engineering, VP Platform/AI, Head of AI Infra, platform strategy, real-time AI, multi-language ASR, speaker identification, enterprise SaaS, GTM alignment, reliability at scale, COGS optimization, vendor negotiations.
• Scaled Otter Pilot and Data Platform from 0 to ~39 engineers/managers across Canada/Seattle/San Diego; built a Senior→Staff→EM pipeline and sustained best-in-class retention (3 departures in 6 years). • Architected and led Otter Pilot, pivoting Otter from mobile-first recording to a cloud meeting assistant; tripled ARR into the high–eight–figure range in year one with rapid virality and seamless join across major conferencing ecosystems. • Commercialized multilanguage ASR (early launches drove low–seven–figure ARR within months, zero marketing) and delivered Real-time Speaker Identification; expanded via a hybrid in-house + partner approach across multiple European and Asian languages with superior in-house performance. • Drove multi-million annualized infrastructure savings and stronger resilience through micro-VMs over K8s, ARM migrations, autoscaling, and packaging changes; significantly reduced incident response times and improved reliability to enterprise-grade levels. • Instituted a 25% Reliability OKR mandate and cross-functional pods; operationalized ML research by pairing Staff/Senior backend with researchers; deepened partnerships with Zoom and Microsoft, unlocked enterprise revenue with a rapid Webex MVP, and secured multi-year AWS discounts by leveraging a credible GCP plan.
• Architected the Virtual Assistant/meeting platform that shifted Otter from mobile-first to a cloud meeting solution—laying the groundwork for Otter Pilot and the company’s remote-first growth. • Engineered horizontal sharding and a hot/cold storage tier (S3-backed) to remove capacity ceilings and unlock effectively unbounded transcription scale. • Eliminated recurring outages in core systems by introducing high-availability patterns and observability—upgrading Redis with monitoring and hot failover and hardening Elasticsearch for resilient search—preserving uninterrupted live transcription while enabling rapid, offline root-cause analysis. • Built the data engineering foundation powering ARR, churn, engagement, and A/B experimentation analytics—compressing analysis cycles from days/weeks to hours and materially increasing Data Science velocity for decision-making. • Led SOC 2 success on the first attempt; embedded privacy-by-design into the data lake (PII tokenization, encryption in motion/at rest), streamlining HIPAA readiness and clean data governance. • Founded and grew two foundational teams (Big Data Platform and PaaS) from 0 to 7; established architecture guardrails, ownership models, and operating rhythms that scaled with product demand. • Credibility: patents filed/awarded during this period in virtual meeting auto-join and real-time conversation summarization.
• Led the real‑time integrity measurement and investigation platform serving Facebook, Instagram, and WhatsApp, enabling safety teams to analyze signals across modalities (text, image, video, graph, behavioral). • Designed and launched a real‑time investigation accelerator that isolates root causes across time, geo, and feature dimensions—compressing investigative cycles from days to minutes and improving decision quality versus seasonality‑sensitive, model‑only signals. • Enforced a rigorous harm‑reduction bar (≥20% impact) for automation; piloted graph‑based correlation to link clusters of harmful content and decisively halted low‑ROI paths to protect compute budgets. • Socialized anomaly‑detection and investigation best practices via internal talks and playbooks, catalyzing wider adoption by integrity data scientists and partner teams. • Managed a 10-person team of engineers and data scientists; served on the monthly SEV committee reviewing SEV‑1/SEV‑2 incidents to align real‑time readiness with executive expectations.
• Acting Software Engineering Manager for AutoFocus; rebuilt and led a 10-person team (7 backend incl. principal, 2 UI/UX, 1 middleware) with 2 DevOps after an org shift, hiring 4 critical early‑career backend engineers to stabilize and scale ingestion. • Established execution rigor—two‑week Agile releases, on‑call rotations, and blameless RCAs/postmortems—while steering incident response and production change during major migrations. • Lifted reliability to 99.99% over 6 months for enterprise customers by eliminating HBase/Elasticsearch brownouts and codifying operational runbooks. • Drove $1M/year in OPEX savings via architectural optimization and a migration plan to a multi‑tenant datacenter outside Silicon Valley (separate from additional hosting savings). • Delivered an active–passive DR strategy and executed the datacenter migration (fall 2018), strengthening business continuity for AutoFocus investigations. • Operated Kafka‑based ingestion pipelines and microservices (Java/Python) with HBase/Cloudera and SQL, processing ~1 TB/day of APT data across private multi‑U.S. datacenters. • Partnered closely with PM, WildFire teams (as upstream data providers), and malware researchers/analysts to enrich WildFire outputs inside AutoFocus and steer the threat intelligence roadmap. • Authored a next‑gen ingestion/search architecture using a columnar store targeting 10x scalability and faster analytics (design completed during this period).
• Founding engineer and later technical lead for AutoFocus, a real‑time threat intelligence system ingesting ~4,000 events/sec (~1 TB/day) at GA and scaling 10x within months; delivered from inception to GA in six months, then on a six‑week release train. • Designed a resilient streaming architecture using Kafka and RMQ feeding custom pipelines backed by HBase (system of record) and Elasticsearch (indexed search), with MapReduce batch backfill; implemented replay semantics delivering zero data loss (RPO ~0) and DC‑level recoverability across private data centers. • Optimized for long‑term retention and analyst concurrency via tiered indexing, schema design, and query planning—enabling rapid investigative pivots across IOCs, samples, campaigns, and malware families without sacrificing throughput. • Advanced analytics with graph enrichment to correlate APT activity to malware families/IOCs; enabled downstream automated protection policies for administrators to block active threats. • Shaped the v2/2.0 roadmap with Product Management and the WildFire team; provided technical leadership across DevOps/SRE, ML, and security research (8–12+ core partners), mentored two new‑grad PhDs and an intern, and hired a mid‑level engineer to scale capacity. • Contributed to the commercial trajectory as AutoFocus matured from a WildFire upsell motion to an independent, multi‑million‑dollar ARR product within two years. • Architected cost‑aware storage patterns supporting WildFire’s long‑term retention requirements, improving reliability posture while managing compute/storage spend. • Recognized with multiple company-wide innovation awards (2015–2017) for AutoFocus, threat intelligence analytics, and security architecture excellence.
• Owned and modernized Palo Alto Networks’ URL Filtering service (subscription add‑on), delivering line‑rate, low‑ms on‑device enforcement with cloud‑backed continuous classification; identified 10M+ malicious web threats daily across a global enterprise customer base. • Cut threat‑intel publishing from 24 hours to 30 minutes (>95% faster) by redesigning the big‑data pipeline with MapReduce/Pig, Solr indexing, and self‑healing automation across AWS/Hadoop—accelerating protection updates to customers in near real time. • Architected and delivered a private, air‑gapped URL Filtering appliance for highly regulated and military‑grade deployments; briefed senior defense leaders and enabled disconnected operations without sacrificing policy fidelity. • Drove cross‑functional delivery as an IC lead, partnering with web‑scale crawling teams and network management to integrate URL categorization into the firewall data plane and cloud services; helped differentiate against legacy ACL‑based firewalls and strengthen subscription attach. • Tackled L7 scale challenges (e.g., thundering herd) and optimized user discovery pathways to sustain high‑growth enterprise traffic at line rate. • Contributed to mobile security via GlobalProtect Mobile Security Manager and implemented certificate‑based authentication/SSL for MDM enablement. • Authored US Patent 10,560,478 (Policy Enforcement): parsing device logs to map IPs to users and apply user‑aware session policies.
• Founded a bootstrapped studio that built and shipped a portfolio of iOS apps (10K+ cumulative downloads, 4.5★ average) and advised five first-time, pre-seed founders from idea to investor-ready MVPs. • uClimateSF: Designed a micro‑climate weather app for San Francisco using Weather Underground and community weather stations; dynamic station updates without app releases; hourly ingestion on private Linux hosts; 100% reliable with no ingestion delays; 45% of total portfolio downloads. • iOnTime: Built a privacy-first calendar + location assistant that computed leave-time on-device (no data shared, no third‑party trackers, no monetization of location); launched ~18 months before Apple’s similar iOS 9 feature; 40% of total portfolio downloads. • PocketChange: Lightweight, multi-currency expense tracker for travelers; offline-first design for quick entry and categorized summaries; 15% of portfolio downloads. • Full-stack ownership end-to-end: product strategy, PRDs/specs, UX, Objective‑C iOS development, Python services on Ubuntu/AWS, analytics/ASO, release management. • Advisory impact: 4–12 week engagements guiding PRD/functional specs, architecture blueprints, and staffing pathways; all five founders produced investor-ready MVPs; consulting revenue ~$45K, app revenue ~$6K.
• Senior IC owning embedded architecture for intelligent lighting at scale—power-line communication, sub-10ms real-time control, and self-healing firmware to meet ceiling-plenum, zero-failure requirements across U.S. enterprise deployments. • Drove 90% energy savings in data centers and ~75% savings vs. fluorescent baselines; marquee adopters included SAP, Volkswagen, Johnson Controls, and Facebook/Microsoft hyperscale facilities. • Secured patent US 20110241567 A1 for a hybrid fluorescent light controller, enabling dimmable retrofits that de-risked customer pilots and opened pathways to new-build contracts without costly rip-and-replace. • Built a modular codebase that scaled from 16KB to 32KB TI MSP430 MCUs while adding I2C and custom 1‑Wire support; delivered three controller SKUs within months, accelerating variant launches and reducing engineering overhead. • Coordinated with hardware, field ops, and customer success for seamless on-site upgrades (e.g., Facebook Prineville), while collaborating with TI/Zilog to optimize sensor and storage integrations across fixtures. • Helped underpin Redwood’s award streak: Architectural SSL PIA (Platinum), Edison “Best New Product: Energy,” Bloomberg NEF Pioneer (2011), and CIBSE Building Performance (2013).