Harshit Vashisth

With 8 years of experience, I am an ambitious and accountable professional, consistently driven by a logical and pragmatic approach to problem-solving. I thrive on autonomy, taking ownership of complex challenges and delivering impactful, data-driven solutions with a structured and disciplined mindset.

Experience

M

Data Science Intern

Medtek.ai · Jan 2026 – May 2026

Not yet confirmed

• Built Clinyx, a clinical AI desktop assistant using Electron, React, Python, and Claude API for clinical programming workflows.
• Designed a parallel multi-agent architecture for automated SDTM, ADaM, and TFL code generation with real-time streaming.
• Delivered a secure PDF-to-spec automation, converting annotated CRFs into SDTM Excel specifications across 8+ CDISC domains.

Data Scientist

MEDTEK AI · Jan 2026 – May 2026

Not yet confirmed

• Built Clinyx, a clinical AI desktop assistant (Electron, React, Python, Claude API) automating clinical programming workflows for biostatistics teams.
• - Designed a parallel multi-agent architecture for automated SDTM, ADaM, and TFL code generation with real-time streaming and orchestration.
• - Built an end-to-end PDF-to-Spec pipeline converting annotated CRFs into SDTM specifications across 8+ CDISC domains using MuPDF, cutting manual effort per study.
• - Engineered secure, enterprise-ready AI workflows with authentication, role-based access, and PostgreSQL.
• - Optimized LLM-powered document processing and backend APIs for high-performance clinical applications.

C

AI Innovation Incubator (AIII)

Concept2Action · Aug 2025 – Oct 2025

Not yet confirmed

• Implemented a multi-agent legal AI assistant using OpenAI, Gemini, Azure AI, and RAG, reducing research time by 95%.
• Developed AI-powered retrieval pipelines processing 250GB+ of legal data with citation-backed search.
• Architected a scalable full-stack platform with FastAPI, React, and LangChain, supporting 50K+ users with 99% uptime..

C

AI Innovation Incubator

Concept 2 Action · Aug 2025 – Oct 2025

Not yet confirmed

• - Cut legal research time 95% with a multi-agent, multi-RAG assistant (OpenAI, Gemini, Azure AI) — grounding every answer in source law for citation-backed results.
• - Scaled full-stack platform (FastAPI, React, LangChain) to 50K+ users at 99% uptime; AI Search processing 250GB+ of legal data.
• - Architected secure retrieval with FastAPI, AWS S3, Redis, OAuth 2.0 for fast, reliable responses.
• - Led full-stack and LLM orchestration (React, TailwindCSS, LangChain) with multilingual support.

A

AI Engineer (Freelance)

Adjibar Tech · Jun 2024 – Jul 2025

Not yet confirmed

• Deployed 10+ ML pipelines on Databricks & Azure ML with Prometheus & ELK, serving 1M+ requests/month.
• Fine-tuned LLMs and RAG models (LLaMA, DeepSeek R1, Hugging Face, OpenAI) for real-world applications like IntelliCore.
• Boosted model accuracy by 12% through team-driven ML solutions and efficient experiment tracking with MLflow & DVC.

AI Engineer

Adjibar · Jun 2024 – Jul 2025

Not yet confirmed

• - Deployed 10+ ML/LLM pipelines on Databricks and Azure ML serving 1M+ requests/month.
• - Built Prometheus + ELK observability and MLflow/DVC tracking; lifted model accuracy 12% through rigorous iteration.

D

Research and Software Engineering Intern

Defense Research and Development Organization (DRDO) · Jun 2022 – Jul 2022

Not yet confirmed

• Engineered Packet Sending Tool for sending and receiving UDP/TCP packets in network testing.
• Crafted a Java-based chatbot recognizing 100+ patterns, enhancing software usability and interaction.
• Optimized a high-throughput traffic generator sending 829K+ packets/minute utilizing multithreading and Java Swing.

D

Summer Intern

Defence Research and Development Organisation (DRDO) · Jun 2022 – Jul 2022

Not yet confirmed

• Created a Packet Sending Tool which can send 832433 Packets in 60sec and with a speed of 1566kb/sec on single thread.