Kaushik Gandhi A on Tal
Headline: AI Development Associate @ MS Clinical Research | Founding AI Engineer | Taking clinical research from legacy to intelligent
About: I'm an AI Engineer with one obsession: if the work is manual and repetitive, it can probably be automated — and it should be. As the founding AI Development Associate at MS Clinical Research, I'm building the AI layer of a clinical research organisation from a clean slate — replacing legacy, manual workflows with intelligent automation and agentic systems powered by Claude and modern LLMs. The brief is simple: take the org from legacy to intelligent. In a short span I've already built a big part of that layer: a unified, monday.com-style operations platform that brings the organisation's separate clinical-research apps under one roof, an AI assistant embedded across the workspace, and a shared design system and common data model so those apps finally behave like one connected system instead of silos. Day to day, that's the job: find the busywork — pipelines, reporting, internal tooling, decision support — and replace it with an AI solution or automation, so people spend their time on the work that genuinely needs a human. I hold an M.Sc. in Artificial Intelligence & Machine Learning (completed May 2026), with research in LLM confidence estimation, hybrid modeling, and scalable MLOps for deployment-ready systems. Tech Stack: Python, PyTorch, TensorFlow, LLMs (Claude & others), RAG, NLP, LangChain, Hugging Face, Vector DBs, MLOps, Docker, MLflow, Airflow, FastAPI, Streamlit, AWS, GCP, SQL, Power BI.
Location: Bengaluru
AI stack:
- claude code
- elevenlabs
- github copilot
- hermes agent
- langchain
- obsidian
Work experience:
- AI Development Associate at MS CLINICAL RESEARCH, Jun 2026 - Present · 3 mos, Bengaluru, I don't automate tasks. I retire them. I walked into a clinical research organisation running on manual workflows and disconnected tools, and I'm building the layer that makes all of it think. A unified operations platform so the company's separate clinical apps finally speak the same language, on one data model and one design system. An AI assistant that lives inside the workspace, so the answer comes to you instead of you going to find it. A project-management system and a CRM being rebuilt from the ground up, templates, timelines, pipelines, invoicing, the busywork disappearing piece by piece. The mandate is to take this place from legacy to intelligent. I didn't wait for permission to start. I'm building it now, and it's already changing how the work gets done. That's the job. I find the work that shouldn't need a human, and I make sure it stops touching one.
- AI Software Engineer(trainee) at Worktual AI, Feb 2026 - Jun 2026 · 5 mos, Chennai, Tamil Nadu, India, - Engineered a centralized multi-agent CRM orchestration platform integrating business-specific tool functions across Sales, Support, Operations, and other organizational departments into a unified application ecosystem. Designed the system to automate and streamline end-to-end customer lifecycle workflows from lead management, deal tracking, customer profiling, scoring, and interaction summaries to cross-department task orchestration significantly reducing manual operational processes and enabling seamless collaboration, data synchronization, and intelligent workflow automation across the organization. - Developed, tested, and validated a multilingual Phone AI Agent supporting Tamil, Telugu, Malayalam, Hindi, and Kannada by building the complete real-time conversational voice pipeline, including audio ingestion, Voice Activity Detection (VAD), Automatic Speech Recognition (ASR), Speech-to-Text (STT) processing, multilingual language detection, LLM-driven conversational orchestration with contextual memory and function calling, response generation, Text-to-Speech (TTS) synthesis, and real-time audio streaming for natural, low-latency, and context-aware voice interactions. -Conducting advanced R&D on Speech-to-Speech (S2S) architecture leveraging the Mimi codec and LLM-based inner monologue frameworks, including model training and inference optimization on NVIDIA H100 GPUs for low-latency conversational AI systems. -Designed and implemented integrations using the Model Context Protocol (MCP) to expose and connect tool functionalities from our products for seamless external consumption across specific business needs. The architecture includes a core tool layer, client component, and server component, enabling scalable, secure, and efficient interoperability between systems. -Established and maintained comprehensive internal AI documentation and knowledge-sharing frameworks, significantly improving cross-team collaboration, feature adoption, and development efficiency.
- Stock Analyst at Decathlon, Oct 2022 - Jan 2023 · 4 mos, Chennai, Stock Analyst Intern Began my internship supporting the sales function, where I was responsible for managing the visual merchandising and operational performance of an entire section. This included optimizing product arrangement, customer engagement, accessibility, touch-and-feel experience, and overall store presentation across multiple merchandising principles. Expanded my responsibilities to include inventory optimization and stock management, ensuring accurate stock movement, replenishment, and product availability while supporting day-to-day retail operations. Was entrusted with revamping an underperforming section with the objective of improving customer engagement and increasing sales. Independently analyzed customer behavior, seasonal demand, product trends, and business priorities to develop and execute a strategic merchandising plan. Successfully transformed the section by implementing data-driven layout improvements and merchandising strategies, resulting in increased sales performance, higher customer walk-ins, and stronger overall engagement. Managed the entire initiative independently, earning the trust of my Team Lead through ownership, strategic thinking, and consistent execution.
- Machine Learning Engineer at Object Automation, Aug 2022 - Oct 2022 · 3 mos, Built and deployed machine learning and AI models using Python, TensorFlow, and PyTorch, focusing on scalable, production-ready ML systems. Performed data cleaning, feature engineering, and exploratory data analysis (EDA) on structured datasets to improve model performance. Implemented and optimized models using TensorFlow and PyTorch, focusing on accuracy, scalability, and inference efficiency. Collaborated with data engineers and software developers to integrate ML models into production-ready pipelines. Monitored post-deployment model performance, identified drift, and supported retraining strategies to maintain accuracy over time.
Education:
- Master of science - AIML, Artificial Intelligence and machine learning, Christ University, Bangalore
- Bca, Computer applications, SRM IST Chennai
Public links:
- github: https://github.com/KaushikGandhi-AIML
- medium: https://medium.com/@kauxhik77
Tal is a social hiring app where bosses and talent meet directly.