FORECAST · ACT
Client engagement
Pharma Copilot Platform
End to end AI platform for pharmaceutical manufacturing with autonomous control loops, demand forecasting, and automated compliance reporting.
- GenAI
- LSTM
- XGBoost
- Q-Learning
Granth Gaurav / AI Engineer
contact →Data engineering and applied AI for regulated industries.
pipeline · 07select a stage
DELHI NCR, INDIA · IST IST
pipeline active ·active ·07
Positioning
What the work actually is.
I build production GenAI and agentic systems for regulated industries. Pharmaceutical manufacturing, quality assurance, enterprise business intelligence.
The work sits inside real constraints. Batch records are auditable. Validation is not optional. A model that hallucinates a specification is a recall.
Day to day that means Python, SQL, LangGraph and LangChain over Databricks and cloud, with reinforcement learning and forecasting where control loops actually earn their keep.
- production
- auditable
- regulated
- agentic
Selected work
Ships to prod.
04 projects
Distracted and Drowsy Driver Detection
Real time safety application detecting driver fatigue, seatbelt compliance, and phone usage. Cross platform Flutter client with a Flask backend for instant alerts.
- YOLOv8
- Dlib
- Flutter
- Flask
ISL Sign Language to Speech Translation
Real time Indian Sign Language translation system at 97 percent accuracy on a custom curated dataset, with CNN layers optimized for low latency inference on edge devices.
- CNN
- TensorFlow
- OpenCV
sshhh-lanchat
Ephemeral encrypted terminal chat for a local network. No server, no accounts, nothing written to disk. Single static binary.
- Go
Timeline
Where the work happened.
Delhi NCR, India
Navikenz
Software Engineer, Data Engineering and AI
GSK Marketing Automation
Spearheaded a GenAI-driven QA/QC framework that tested marketing campaign files for correctness before market release, reducing manual effort and operational costs by 80 percent through automated validation on Databricks.
GenBI Text-to-SQL
Developed a text to SQL conversion engine leveraging LLMs to enable data retrieval and automated visualization for non technical stakeholders.
Delhi NCR, India
Navikenz
Software Engineering Intern
AI and Cloud Training
Completed structured training across traditional machine learning, deep learning, generative AI, Azure cloud services, FastAPI, and SQL.
Capstone Project
Built Pharma Copilot as the training capstone: an intelligent pharmaceutical-manufacturing system using reinforcement learning and forecasting models to improve production efficiency and help prevent batch defects.
Kharagpur, India · On-site
Indian Institute of Technology, Kharagpur
Research Intern
IoT and Cloud
Led projects integrating IoT devices with cloud platforms to improve operational efficiency.
Django and AWS
Developed scalable applications with Django and AWS, strengthening server-side performance and reliability.
Systems Practice
Applied Python, system architecture design, and cloud-computing fundamentals across research projects.
Remote, Chandigarh
Intel Corporation
Data Science Intern
Sign to Speech Translation
Researched and implemented a sign language to speech engine using transformer architectures and computer vision.
Stack
Intel OpenVINO, Tableau, and Orange for model optimization and data storytelling.
Capability matrix
Tools, by stage.
Columns are stages of a real pipeline. A mark means that tool runs at that stage.
| tool | INGEST | TRANSFORM | MODEL | SERVE | OBSERVE | span |
|---|---|---|---|---|---|---|
| Python | Python used at INGEST | Python used at TRANSFORM | Python used at MODEL | Python used at SERVE | not used | 4 / 5 |
| SQL | SQL used at INGEST | SQL used at TRANSFORM | not used | not used | SQL used at OBSERVE | 3 / 5 |
| PySpark | not used | PySpark used at TRANSFORM | not used | not used | not used | 1 / 5 |
| Databricks | Databricks used at INGEST | Databricks used at TRANSFORM | not used | not used | Databricks used at OBSERVE | 3 / 5 |
| AWS | AWS used at INGEST | not used | not used | AWS used at SERVE | not used | 2 / 5 |
| Azure | Azure used at INGEST | not used | not used | Azure used at SERVE | not used | 2 / 5 |
| Docker | not used | not used | not used | Docker used at SERVE | not used | 1 / 5 |
| PyTorch | not used | not used | PyTorch used at MODEL | not used | not used | 1 / 5 |
| TensorFlow | not used | not used | TensorFlow used at MODEL | not used | not used | 1 / 5 |
| LangChain | not used | LangChain used at TRANSFORM | not used | not used | not used | 1 / 5 |
| LangGraph | not used | not used | LangGraph used at MODEL | LangGraph used at SERVE | not used | 2 / 5 |
| OpenCV | not used | not used | OpenCV used at MODEL | not used | not used | 1 / 5 |
| Power BI | not used | not used | not used | Power BI used at SERVE | Power BI used at OBSERVE | 2 / 5 |
| Git | not used | not used | not used | not used | Git used at OBSERVE | 1 / 5 |
- Languages
- Python · SQL · JavaScript
- AI and ML
- Generative AI · agentic systems · computer vision · deep learning · reinforcement learning · NLP
- Cloud and platforms
- AWS · Azure · Databricks · Git · Docker
Credentials
On the record.
Certifications and awards
2025
Databricks Certified Data Engineer Associate
verify credential, opens in a new tab2025
Oracle Generative AI Professional
verify credential, opens in a new tab2024
Best Project Award, Chandigarh University
Driver Activity Monitoring System
2024
NPTEL Star Performer, top 2 percent nationally
IoT course, IIT Madras
2022
1st Runner-up, AirOthon
US Consulate Kolkata and KIIT-TBI
Education
Chandigarh University
Jun 2025
B.E. Computer Science and Engineering
GPA 8.28 / 10.0 · Chandigarh, India
Delhi Public School
Jun 2021
Class XII, CBSE, PCM
91 percent · Bokaro Steel City, India
Delhi Public School
Jun 2019
Class X, CBSE
92.4 percent · Bokaro Steel City, India
Publications
A Holistic Approach to Detecting Distracted and Drowsy Drivers
Apr 2025
Research on a multimodal driver-monitoring system for distraction, drowsiness, phone use, and seatbelt compliance.
read on IEEE, opens in a new tab