Jan 2025 – Present
Sazim Tech — Software Engineer (L4)
Two client engagements at once: an AI spreadsheet editor at ProposalAI, and high-concurrency APIs and multimodal inference at Perle AI. Mentoring six engineers alongside the client work.
Software engineer at Sazim Tech Ltd., working remotely across two client engagements and the consultancy's own internal work.
Client: ProposalAI (Toronto, Ontario, Canada)
Owned and shipped an AI spreadsheet editor generating thousands of rows (React, NestJS, Phoenix, FastAPI)
Cut large-spreadsheet load latency 9x by compressing payloads 105x with gzip and custom data transformers
Architected a high-recall GenAI data pipeline using LanceDB and PostgreSQL, reducing model hallucination to 0.5%
Offloaded bulk generation to BullMQ and Redis workers, keeping the editor responsive during long-running LLM jobs
Built unit and end-to-end coverage with Vitest and Playwright in GitHub Actions, catching regressions before release
Client: Perle AI (San Francisco, California, US)
Built secure, high-concurrency REST APIs and backend matching services in Node.js/PostgreSQL serving 10,000+ active users with sub-100ms p95 response times
Cut manual review 10 hours/week with an LLM-as-judge evaluation service in Python (OpenAI SDK, PostgreSQL)
Built a WebSocket sync layer that pushed updates to users instantly, removing the 30-second refresh delay
Sustained 99.9% uptime on a multimodal inference pipeline serving 100,000+ image, audio, and text files, built with FastAPI and Redis task queues
Internal
Mentored 6 engineers and led 2 teams for both internal and client projects
Built internal tools, scoped requirements with client stakeholders, and interviewed candidates for engineering roles