Role & Responsibilities

Role

AI/ML Intern at SynerSense Pvt. Ltd.

Responsibilities

1) AI / ML (Primary Focus)

• Design and refine vision-language model workflows, including prompt templates, role/system instructions, and structured output schemas for stable downstream parsing.
• Build repeatable evaluation pipelines for model behavior analysis: prompt determinism checks, float/score consistency validation, and trial-to-trial variance tracking.
• Run experiment cycles for hyperparameter and prompt optimization, comparing trial artifacts to identify robust configurations instead of one-off best runs.
• Implement model-promotion logic from best-trial outputs into final training/inference paths, with reproducibility checks on input/output contracts.
• Translate research findings into production-oriented decisions by balancing model quality, latency, and operational reliability.

2) Deployment & MLOps (Secondary Focus)

• Containerize AI workloads using Docker with explicit runtime parity controls (base-image pinning, engine compatibility, dependency lockfile integrity).
• Deploy and operate cloud execution paths on Hugging Face Spaces/Jobs, including remote run monitoring, log polling, and artifact synchronization.
• Diagnose and resolve cross-environment failures (local vs cloud) such as Node/pnpm engine mismatches, dependency resolution conflicts, and runtime startup errors.
• Implement observability features for long-running jobs: status APIs, log-streaming endpoints, and structured metadata capture for debugging and audits.
• Strengthen release reliability through CI-integrated validation (navigation checks, content-quality checks, generated-doc consistency, and reproducible site builds).

3) Fullstack Engineering (Supporting Focus)

• Build backend services with FastAPI for experiment/job lifecycle control (start, list, status, logs, stop), including safer process cleanup and failure-state handling.
• Design API contracts that support asynchronous workloads, incremental log delivery, and consistent response schemas for frontend consumption.
• Develop Svelte/SvelteKit interfaces for real-time operational visibility: live logs, trial comparisons, version-aware exploration, and model-result inspection.
• Connect frontend and backend with reliability-first patterns (defensive state handling, retry-aware polling flows, and graceful empty/error states).
• Improve end-to-end system quality by aligning data contracts, API semantics, and UX behavior across model experimentation and deployment workflows.

Acknowledgments

I extend sincere gratitude to Praveen Kulkarni Sir for his guidance, encouragement, and insights throughout this journey. I also appreciate the freedom and trust given by the SynerSense team to explore creative solutions and contribute meaningfully to the pipeline.