Md Sadman Sakib
Software Engineer | On-Device AI, ChromeOS & Backend Systems
Software engineer with 6+ years of experience across AI product features, ChromeOS systems, and backend infrastructure, with recent Google ownership in privacy-preserving local intelligence and prior high-scale performance optimization work.
Professional Experience
Google
L&T ODML (On-Device Machine Learning)
- Led near-instant semantic clustering and retrieval for local files on Googlebook, enabling privacy-preserving file organization and downstream first-party AI experiences. Tech: AICore, Kotlin, small language models, embedding models.
- Contributed to an on-device Context Engine for locally stored files, generating structured context for product integrations. Tech: AICore, LLMs, LoRA fine-tuning.
- Led end-to-end development of Expand Background, the first on-device Imagen-powered generative AI feature for Chromebook Plus, expanding images locally under 2s and featured in a public Google press briefing.
- Automated text-image alignment evaluation for Imagen model training, reducing manual engineering effort by 90%.
ChromeOS Bisector
- Redesigned pub/sub flow control, reducing traffic by 80%, eliminating idle VM runners, and reducing GCP operational cost by $5k+/year.
ChromeOS HW Security
- Hardened TPM-backed credential security by mitigating SSD wear-leveling data remanence; decoupled OOBE/Login services and deprecated 6,000+ lines of legacy code.
- Developed an Integration Test Coverage profiling tool adopted by the team to guide coverage improvements and bug detection.
Penta Global Ltd.
- Achieved a ~1500x performance increase by redesigning a legacy service handling 300M+ records with table partitioning, indexing, and Redis caching.
- Developed a high-concurrency ticketing service using Java Spring Boot and Redis distributed locking to aggregate complex cross-service responses.
- Designed a greedy Query Manager algorithm to optimize range queries and maximize cache hit ratios across distributed datastores.
Samsung R&D Institute, Bangladesh
- Developed native macOS features for Samsung DeX in Objective-C and built Java-based IoT test automation tools for hardware validation workflows.
Technical Skills
- Machine Learning: On-device AI, AICore, embeddings, LLMs, LoRA fine-tuning, Computer Vision, LangChain, LangGraph
- Backend & Systems: FastAPI, Spring Boot, Microservices, Redis, SQLite, PostgreSQL, distributed locking, caching
- Languages & Tools: C++, Go, Python, Java, Kotlin, Git, Linux, Docker, Playwright, FFmpeg, Claude Code, Gemini CLI
Education & Research
- B.Sc. in Computer Science & Engineering: University of Dhaka, 01/2015 - 12/2018; degree received 08/2019.
- ICLR 2019 Workshop: Co-authored A Gray Box Interpretable Visual Debugging Approach for Deep Sequence Learning Models, introducing d-DeVIS for visual debugging of audio-based models. arXiv:1811.08374
- Cognitia OSS: Solo architected a FastAPI/React multimodal AI evidence platform with Redis/RQ, Gemini, ffmpeg, Playwright, and systemd; built hybrid video sampling using scene cuts, motion spikes, periodic coverage, dedupe/downsampling, and embedding-backed insight search; covered by 50+ tests.
Awards & Leadership
- ICPC Dhaka Site: 10th among 298 teams; NCPC: 17th among 150 teams; Codeforces: reverse_macro, max rating 1811.
- Champion: NSU Power Buzz Programming Contest 2018.
- Trainer - DevSkill: Trained and mentored 500+ students in data structures, algorithms, and competitive programming.
- Problem Setter - InterviewKickstart: Set, developed, and reviewed 50+ technical interview problems.
- Co-Founder & Editor - Shoshikkha.com: Built Bengali CS learning content platform.