Deepam Shah
Senior Software Engineer who builds backends that bend without breaking — founding engineer on Zoom Workplace for Frontline, ex-Goldman Sachs. Now building production LLM & agentic systems and pursuing a Ph.D. in AI.
↑ a real terminal — type help, or tap a command
~/about $
I build backend systems that don't blink under pressure. As a founding engineer on Zoom Workplace for Frontline, I took the core backend from zero to a product serving 45K+ frontline workers across 120+ enterprise accounts — then shipped RAG and agentic AI features on top of it. Earlier, at Goldman Sachs, I cut SQL latency 35% and debug time 60%.
Now I'm pushing into the next frontier — pursuing a Ph.D. in Artificial Intelligence while engineering full-time, exploring how LLMs, retrieval, and agents reshape the systems we build. When I'm not shipping production code, I'm prototyping AI side projects — this site is one of them.
~/experience $
Senior Software Engineer — Founding Engineer, Zoom Workplace for Frontline
06/2022 — PresentZoom Communications
- Founding backend engineer for Zoom Workplace for Frontline, scaling a multi-tenant workforce platform from 0→1 to 45K+ frontline workers across 120+ enterprise customers in retail, healthcare, and manufacturing. Co-architected the backend (Java 17, Spring Boot 3, MySQL, Redis, DynamoDB, Kafka) and led backend development across most of the product — shift scheduling, shift channels, swap/cover workflows, task management, and on-call operations.
- Designed and delivered flagship Frontline capabilities end-to-end — Persistent Shift Channels (always-on Team Chat + Push-to-Talk surviving shift transitions, backward-compatible across 3M+ existing channels) and Kafka event-driven services for channel lifecycle, notification fan-out, and bulk operations — with CSV import pipelines onboarding 5K+ users per upload that cut enterprise setup from ~4 hours to under 10 minutes (~95% faster).
- Shipped a steady stream of recent platform capabilities: geo-fenced punch-in integrated with UKG Pro WFM, SAML/SSO role mapping for automated shift-group channel assignment, and a QR-code notification system (scan-to-notify shift groups for help requests and issue reporting).
- Delivered AI-powered collaboration — RAG shift summaries, natural-language Q&A over workforce data, AI channel summaries, and AI-assisted message composition — on an LLM orchestration layer (guardrails, function calling) serving thousands of requests/min at sub-second p95 latency.
- Hardened the platform with JWT, Spring Security 6, iOS App Attest, Android Play Integrity, PII masking, and shift-group audit logging; remediated 200+ vulnerabilities flagged by Mend, Black Duck, and Prisma.
- Drove quality as a technical lead: raised automated test coverage from 45% to 85%+, cut customer-reported production defects ~40% QoQ, and mentored 5+ engineers through design and code reviews.
- Earlier in tenure: built backend for Zoom Huddles (workspace management, participant state sync, real-time collaboration); architected the Google Drive and ServiceNow Marketplace apps (OAuth 2.0, bidirectional APIs, deep-link routing); and built the Slack→Zoom Migration Platform with fault-tolerant ETL for channels, users, message history, file assets, and permissions.
Java 17Spring Boot 3KafkaAI CompanionRAGMySQLRedisSpring SecuritySoftware Developer Consultant
06/2021 — 06/2022Goldman Sachs
- Built and tested production REST APIs (Java 11 / Spring Boot) for core lending modules on GS Select; cut SQL latency 35% on high-traffic endpoints.
- Led migration of loan-servicing microservices from Java 8 to Java 11 (Spring Boot, Gradle, Atlas), improving throughput by 22% while reducing technical debt.
- Engineered savings-servicing money-movement features with functional-level JUnit coverage, and built an async auditing framework that traced the full loan lifecycle — reducing production debug time by 60%.
- Resolved 10+ high-priority tickets and mitigated customer-impacting incidents on-call, sustaining SLA compliance.
Java 11Spring BootSQLMicroservicesJUnitSoftware Developer Intern
06/2020 — 08/2020Esri Inc.
- Prototyped a Git-style version-control system for Jupyter notebooks (save/revert/rollback) with headless kernel access via REST through the Jupyter kernel gateway.
- Containerized ArcGIS Notebook Server on Windows Docker, and applied geospatial deep learning with ArcGIS for feature extraction on aerial/satellite imagery.
PythonDockerDeep LearningGISSoftware Engineer Intern
05/2019 — 08/2019TraceLink Inc.
- Built a Java tool that auto-generates codebase documentation nightly — adopted company-wide across all team locations.
- Automated the Android test suite with Jenkins + Appium in Docker, increasing QA productivity by 40%.
JavaJenkinsDockerAppiumSoftware Engineer
09/2017 — 07/2018Apexon (formerly Infostretch)
- Delivered multiple concurrent software projects end to end — from design through deployment — using Agile/Scrum.
JavaAgile
~/timeline $
- 2017Started as a Software Engineer at Apexon — shipping projects end to end.
- 2018Began M.S. in Computer Science at RIT.
- 2019Interned at TraceLink — built a company-wide doc generator.
- 2020Interned at Esri — Jupyter version control + geospatial deep learning. Finished M.S.
- 2021Joined Goldman Sachs — REST APIs and microservices for GS Select lending.
- 2022Joined Zoom as a founding engineer on Zoom Workplace for Frontline.
- 2024Completed an MBA (3.9 GPA) at Campbellsville University.
- 2026Started a Ph.D. in Artificial Intelligence — going deep on applied AI.
~/projects $
Filter by stack or search. 3 of 3 shown.
Portfolio Management Tool + Dividend Tracker
Behind loginAn AI-powered portfolio app that aggregates holdings, tracks dividend income and payout schedules, and surfaces LLM-driven insights and rebalancing suggestions — with automated ex-dividend/pay-date reminders, yield-on-cost analytics, and projected annual-income forecasting. Live on this site behind login.
This Portfolio Site
SourceAn AI-assisted build: a public résumé + interactive terminal, with a gated personal dashboard ("the mancave") behind Supabase auth. Shipped to Vercel on a $0 stack — this site is itself one of the side projects.
Language Classifier — Decision Tree + AdaBoost
SourceAn ML classifier distinguishing English vs. Dutch text using entropy/information-gain decision trees (98% accuracy), plus a boosted ensemble of decision stumps via AdaBoost (95% accuracy).
~/education $
- Ph.D. in Artificial IntelligenceUniversity of the Cumberlands2026 — 2029In progress
- Master of Business Administration (MBA)Campbellsville University2024 — 20253.9 GPA
- Master's in Computer ScienceRochester Institute of Technology2018 — 20203.46 GPA
- Bachelor's in Information & Communication TechnologyAhmedabad University2013 — 2017