Yashas R.

⭐⭐⭐⭐⭐ 5.0/5.0 Ex-Amazon, Ex-Adobe Top 1% Expert-Vetted on Upwork —. 150+ AI projects shipped at 5/5 stars and 100% Job Success. Advisor to venture-funded AI startups. I architect production LLM agents, RAG systems, and voice AI for teams that need senior thinking and staffing. 𝗪𝗛𝗔𝗧 𝗜 𝗗𝗘𝗦𝗜𝗚𝗡 ✅ Multi-agent systems on LangGraph, OpenAI Agents SDK, and Google ADK — with MCP and A2A interop for tool use, memory, and cross-framework orchestration ✅ Production RAG with hybrid search, Voyage rerank-2.5 / Cohere Rerank v3.5, agentic retrieval, citation enforcement, and RAGAS evaluation ✅ Voice AI agents on LiveKit Agents + Pipecat with Deepgram Nova-3, Cartesia Sonic-3, ElevenLabs Flash v2.5 — sub-300ms end-to-end latency ✅ LLM fine-tuning and adapter training (LoRA, QLoRA, DoRA) with Unsloth, Axolotl, TRL on Llama 4, Qwen3, Mistral, DeepSeek ✅ Time-series and forecasting systems (TimeGPT, PatchTST, Prophet) for trading, demand, and operations ✅ Computer vision and generative imaging/video (SAM 3, YOLO26, Flux, Veo 3.1, Kling 3.0, Runway Gen-4.5) 𝟮𝟬𝟮𝟲 𝗦𝗧𝗔𝗖𝗞 GPT-5.5, Claude Opus 4.7 / Sonnet 4.6, Gemini 3 Pro, Llama 4, DeepSeek V3.2 · LangGraph, OpenAI Agents SDK, Google ADK, Pydantic AI, DSPy · LlamaIndex, Haystack · pgvector, Qdrant, Pinecone, Weaviate, Milvus · LangSmith, LangFuse, Arize Phoenix · vLLM, SGLang for self-hosted inference · AWS Bedrock, Azure AI Foundry, Vertex AI · Python, FastAPI, Node.js, TypeScript, Next.js · MCP servers and A2A agent communication 𝗕𝗘𝗬𝗢𝗡𝗗 𝗧𝗛𝗘 𝗖𝗢𝗗𝗘 I bring senior product and people leadership — having managed engineering organizations approaching a thousand people and led products from idea to end-of-life. I speak on AI and entrepreneurship at conferences and podcasts, and advise venture-funded AI startups on architecture, hiring, and go-to-market. 𝗖𝗥𝗘𝗗𝗘𝗡𝗧𝗜𝗔𝗟𝗦 ▸ Top 1% on Upwork — Expert-Vetted, Top Rated Plus, 100% Job Success Score ▸ 150+ AI projects shipped at 5/5 stars ▸ Ex-Amazon, Ex-Adobe, Ex-Accenture ▸ Mentor and advisor at multiple venture-funded AI startups ▸ Speaker and panelist on AI and entrepreneurship ▸ Founding partner at a boutique AI engineering studio — Frankfurt and US team 𝗛𝗢𝗪 𝗜 𝗪𝗢𝗥𝗞 Senior engineering and architecture only. Engagements typically begin with a paid 1–2 week discovery sprint to scope architecture, model selection, and milestones — then fixed-scope delivery with weekly demos. I work async-friendly across EU and US time zones and turn down work that doesn't merit senior attention. If you're shipping AI to production and need an architect who's done it 150+ times, message me with what you're building.