Sudhanshu Patil

As an AI Solutions Developer, I am driven by a deep curiosity and an ambitious pursuit of mastery, constantly seeking to understand and implement cutting-edge solutions. I thrive on the autonomy to independently tackle complex challenges, always focusing on pragmatic approaches that deliver tangible impact.

Experience

S

AI Solutions Developer

SimpleWorks · Aug 2025 – Jul 2026

Not yet confirmed

• 𝗠𝗖𝗣 & 𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗜𝗻𝗳𝗿𝗮𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲:
• • Architected a cloud-hosted FastMCP server exposing 10+ tools across Claude, Gemini CLI, and internal apps; integrated MCP as a query mode with 4 tool types across 9 CRM modules with configurable 40k-token memory context.
• 𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗥𝗔𝗚:
• • Built and QA-tested an Agentic RAG framework featuring an Orchestrator, Intent Recogniser, Dynamic Query Splitting, Dynamic Library Switching, and a Mathematical Execution Sandbox with HITL query narrowing using a Hub-and-Spoke architecture.
• • Implemented BOI HITL usecase with multi-traversal response generation (up to 2 HITL levels) and automated session relogin via bash script for extended testing.
• 𝗣𝗲𝗿𝗳𝗼𝗿𝗺𝗮𝗻𝗰𝗲 & 𝗥𝗲𝗹𝗶𝗮𝗯𝗶𝗹𝗶𝘁𝘆:
• • Reduced PII redaction latency by 94% (8s → 0.5s) replacing a spaCy transformer with NER DistilBERT; built triple-verification redaction with guardrails blocking excluded fields from LLM payload and resolved session contamination edge cases.
• 𝗣𝗹𝗮𝘁𝗳𝗼𝗿𝗺 𝗠𝗶𝗴𝗿𝗮𝘁𝗶𝗼𝗻𝘀 & 𝗗𝗲𝘃𝗢𝗽𝘀:
• • Led Python 3.11 → 3.13 + LlamaIndex upgrades on QA for GPT-5 series support (Azure AI Foundry); delivered LlamaIndex-only Production upgrade for PSB with offline wheel packaging for air-gapped banking systems.
• • Executed Dev → QA migrations independently in under an hour, including DB, env, and branch changes.
• 𝗣𝗿𝗼𝗱𝘂𝗰𝘁 & 𝗣𝗹𝗮𝘁𝗳𝗼𝗿𝗺 𝗗𝗲𝘀𝗶𝗴𝗻:
• • Designed end-to-end MVP architecture and UX for R-YaBot: Connector Cockpit, multi-DB connectors, @Mention cross-mode queries, Math Function Registry, RBAC, LangFuse traceability, and health monitoring.
• 𝗘𝗻𝘁𝗲𝗿𝗽𝗿𝗶𝘀𝗲 𝗜𝗻𝘁𝗲𝗴𝗿𝗮𝘁𝗶𝗼𝗻𝘀 & 𝗗𝗲𝗹𝗶𝘃𝗲𝗿𝘆:
• • Integrated Apollo API for one-click CRM contact and account creation from chatbot responses; deployed PLUK (PruLife UK) email triaging model to production.
• • Delivered 8,000+ lines of production code; resolved 5+ major production SQL bugs and 20+ QA defects.

S

AI Solutions Developer

SimpleWorks · Aug 2025 – Jun 2026

Not yet confirmed

• Delivered QA-passed agentic reasoning for BFSI clients including Bank of India, as measured by multi-traversal responses with up to 2 HITL levels, by building an Agentic RAG framework (Orchestrator, Intent Recogniser, Dynamic Query Splitting, Math Sandbox) on a Hub-and-Spoke architecture.
• Enabled agentic tool-use across Claude, Gemini CLI, and internal apps, as measured by 10+ production MCP tools across 9 CRM modules with configurable 40k-token memory, by architecting a cloud-hosted FastMCP server as a native query mode.
• Cut PII redaction latency by 94% from 8 seconds to 0.5 seconds by replacing a spacy transformer with a fine-tuned NER DistilBERT and adding triple-verification guardrails that block excluded fields from every LLM payload.
• Unblocked GPT-5-series adoption, as measured by zero-rollback upgrades on QA and PSB Production, by solo-leading Python 3.11 to 3.13 and LlamaIndex migrations on Azure AI Foundry, including offline wheel packaging for air-gapped banking systems.
• Sustained production reliability as a team of one, as measured by 8,000+ lines of production code, 5+ major SQL fixes, and 20+ QA defects resolved, by owning architecture through development and deployment.

Associate Data Scientist (Intern)

Novartis · Jan 2025 – Jul 2025

Not yet confirmed

• Kept sensitive clinical data fully on-prem, as measured by R-code responses of OSS-comparable quality with zero data egress, by building an on-prem LLM chatbot trained on internal R-coding documentation.
• Eliminated manual intervention for 60+ Data Managers, by shipping VBA and Office Scripts automations with auto-filing and input guardrails, and replacing SAS reporting with an interactive RShiny dashboard.

Associate Data Scientist

Novartis · Jan 2025 – Jul 2025

Not yet confirmed

• > Awarded a 𝗣𝗿𝗲-𝗣𝗹𝗮𝗰𝗲𝗺𝗲𝗻𝘁 𝗢𝗳𝗳𝗲𝗿 (𝗣𝗣𝗢) for performance — declined to join SimpleWorks full-time.
• • Built an on-prem 𝗟𝗟𝗠 chatbot trained on internal R-coding documentation; the model consistently responded in 𝗥 without explicit prompting, with quality comparable to open-source LLMs, while keeping all data within the organization.
• • Automated clinical onboarding workflows for 𝟲𝟬+ Data Managers using 𝗩𝗕𝗔 and Office Scripts — auto-filing, input guardrails, and data filtering steps — eliminating manual intervention in the data collection pipeline.
• • Developed an interactive 𝗥𝗦𝗵𝗶𝗻𝘆 dashboard for clinical data visualization and aggregation, replacing 𝗦𝗔𝗦-based reporting for non-technical stakeholders.
• • Contributed to 𝗘𝗧𝗟 pipeline development supporting upcoming evolutionary data infrastructure changes at Novartis.

S

Summer Intern

SimpleWorks · May 2023 – Jul 2023

Not yet confirmed

What drives their work

The Builder

Turns concepts into functional, deployed systems.

Sudhanshu's superpower

You excel at taking an idea or a problem and building a complete, working solution from the ground up.

How Sudhanshu works

You would excel in a team that values end-to-end ownership and hands-on contribution to developing innovative AI products.