Sriya

Built a multi-agent project-intake and planning system on LangGraph (5-node pipeline: CEO, CTO, Product Manager, Developer, Client Manager) with forced tool-calling against Groq-hosted open-weight LLMs, and shipped a Streamlit UI for end-to-end submission. Delivered a production Next.js / TypeScript marketing website (mumbaimeridians.com) built on a WordPress CMS for a client sports franchise; ran a technical SEO audit via GSC crawl-stats analysis, prioritizing critical fixes (missing robots.txt/sitemap, canonical tags, domain split).

AI Engineer Intern at Zyflux LLP (2026-01 – Present)

• Built a multi-agent project-intake and planning system on LangGraph (5-node pipeline: CEO, CTO, Product Manager, Developer, Client Manager) with forced tool-calling against Groq-hosted open-weight LLMs, and shipped a Streamlit UI for end-to-end submission.
• Delivered a production Next.js / TypeScript marketing website (mumbaimeridians.com) built on a WordPress CMS for a client sports franchise; ran a technical SEO audit via GSC crawl-stats analysis, prioritizing critical fixes (missing robots.txt/sitemap, canonical tags, domain split).

AI Engineer Intern at LedgerPe (2025-08 – 2026-01)

• Architected a Routing Intelligence Agent prototype (FastAPI) ingesting synthetic multi-provider payment telemetry across 36 providers / 6 rails, dynamically re-weighting routing decisions via a pattern-detection engine (two-proportion z-tests, effect-size floor, confidence hysteresis).
• Built a deterministic explainability layer generating auditor-ready rationale per routing decision (candidate ranking, exclusion reasons, per-dimension scores) instead of an opaque LLM-generated explanation, verified with a 13-test pytest suite.
• Developed LeadFlow Agent, a self-hosted LangChain / LangGraph lead-generation agent with an OpenAI-compatible LLM gateway and sandboxed CLI tool execution (Twitter/X, LinkedIn via MCP servers) hardened against command injection, with markdown-based persistent memory.

GenAI Engineer Intern at StartCoding AI (2025-06 – 2025-07)

• Built MVP for a GenAI-powered text-to-video platform (similar to InVideo AI) using RunPod AI, MoviePy, OpenAI API script generation, and Qwen 2.5 VL + FAISS-GPU embeddings for timestamp-based clip generation.
• Delivered production-ready MVP that reduced manual editing effort by 70% and cut video generation time-to-first-draft by 60%.
• Built a multimodal web scraper combining text and visual page extraction to source reference content for automated video generation.

ML Engineer at Advocate Voice (2024-12 – 2025-01)

Building Offline Speech-to-Text App

• Developed an offline speech-to-text app supporting multilingual transcription (English, Hindi, Marathi) achieving 90% accuracy, using Real-time STT and Whisper Live for 500ms-latency transcription.
• Enabled CPU and GPU inference for small, medium, and large Whisper models with CUDA, reducing inference time by 60% on GPU.

Bachelor of Engineering in AI & ML