Sailalith S.Private profile: Upwork shows its earnings only to signed-in members, so we check it every few weeks, not daily. MRR here is a rough estimate (Upwork's six-month earnings ÷ months active). Last read Nov 5, 2026.

I build AI agents that teams actually use at work: multi-agent systems, assistants that answer questions from company data, document-processing pipelines, and the integrations that connect them to the tools teams already run. 13+ years in data science and AI, from leading an AI Centre of Excellence to hands-on agent development. I still write the code. WHAT I BUILD • Multi-agent systems and agent orchestration, with evaluation, guardrails and cost control built in • Conversational analytics: plain-English questions answered with SQL, analysis and charts, with access control over what each user can see • MCP servers and coding-agent tooling that connect Claude, Codex or Copilot to your data and internal systems • Microsoft AI: Copilot Studio and M365 Copilot agents, Azure AI Search, Power Platform, Power BI • Document AI: extraction from messy, scanned and handwritten documents, with validation, audit trails and system integration • Production ML and MLOps, including moving workloads to self-hosted or open-weight models to cut cost • Fractional AI lead: strategy, roadmaps and architecture reviews of existing agent and RAG systems SELECTED WORK • Led the AI Centre of Excellence at a national broadband provider: AI strategy and a two-year roadmap, root-cause-analysis agents for customer retention, internal LLM assistants, and a lower-cost speech-analytics pipeline on self-hosted transcription. I also built its multi-agent conversational analytics assistant: business questions turned into SQL, Python analysis in a sandbox and charts, inside a chat app with single sign-on and table-level permissions. • Clinical paperwork automation for a US health-tech startup: OCR and template matching for lab requisitions, a real-time insurance-eligibility service, HL7 export to lab systems, and an evaluation harness on a labelled test set, in a HIPAA-regulated environment. • Codebase Cortex (open source, PyPI): keeps documentation in step with the code. A multi-step agent pipeline, code-aware retrieval, and an MCP server that exposes its tools to coding agents. • FlowLens and Halcyon (open source, npm): a map and health check for AI-built codebases, and one calm canvas for every AI coding session. • Motion Studio (open source): turns a brief into code-driven motion graphics rendered to MP4. • Earlier: an analytics product that won two Malaysia Technology Excellence Awards (2019), telecom and marketing analytics for brands in Malaysia and Cambodia, and pricing models for EU and LATAM markets. HOW I WORK • Outcome first: we agree what the system must do, and how we'll measure it, before I write code • Working software early, with a demo every week • Evaluations, logging, tests and documentation as standard, plus a handover your team can run • Straight answers, including when something isn't worth building WAYS TO START • Architecture review of a live agent or RAG system, with a prioritised fix list (fixed price) • Take one agent from prototype to production (fixed price or hourly) • Ongoing work as your AI engineer or fractional AI lead STACK Python, TypeScript, FastAPI, Next.js · LangGraph, Google ADK, LangChain, LiteLLM, MCP, A2A · Claude, Gemini, open-weight models · PostgreSQL, ClickHouse, Redis, FAISS, Neo4j · Docker, MLflow, ONNX, Terraform · AWS, Azure, GCP · Copilot Studio, Power BI Based in India (IST); I work comfortably across Europe, Asia-Pacific and US hours. Case studies and demos for these projects are in my portfolio below. Send me what you're building and where it's stuck, and I'll tell you what I'd do first.