200K+
Learners
Concept-first teaching that helps learners move from basics to confidence.
Parampreet Singh
Parampreet Singh
Parampreet Singh
ResumeBased in India
ResumeI'm an AI Engineer focused on building production-grade AI systems, fine-tuning small language models, and scaling Gurmat Darbar, a full-stack platform for the Sikh community. My work sits where deep learning, agentic workflows, and scalable backend architecture meet.
Beyond products, I build communities and learning spaces: organizing PyDelhi meetups, hosting Codex events and hackathons in New Delhi, and teaching Python and machine learning through 70+ live sessions. I care about turning complex ideas into useful tools, confidence, and momentum.
BS in Data Science and Applications, IIT Madras (2022-Present)
LLMs, Mathematical Foundations for GenAI, and Software Engineering.
I build communities, organize meetups and hackathons, and create spaces where people learn and ship together.

Python, machine learning, revision marathons, and the kind of teaching that turns exam fear into momentum.



@Param3021
Subscribers: 4.1K
200K+
Learners
Concept-first teaching that helps learners move from basics to confidence.
70+
Sessions
Regular live sessions covering Python, Machine Learning, and revision support.
500+
Feedbacks
Real learner appreciation from doubt-solving, project guidance, and mentorship.
Wall of Fame


Discover samagams happening near you!
A digital ecosystem for Sikh community events, bringing trusted discovery, community contributions, and thoughtful technology into one useful home.
Visit gurmatdarbar.com
1,500+
Users
300+
Samagams
500+
Contributions
A multilingual, Multi-RAG workflow that understands uploaded posters, extracts event information, and automatically fills structured details so listings are ready to publish.
Built a Sikh community event platform used by 1,500+ people, covering 300+ samagams with 500+ community contributions.
Created an AI poster-intelligence pipeline that extracts multilingual event fields and prepares listings for publication.
Independent Client Work
Jun 2026 - Aug 2026
Built a dashboard and three Indian-dialect voice agents for OT readiness, care follow-ups, and appointment booking using Bolna, LiveKit, Cartesia, and Deepgram.
Designed graph-based agent reasoning and an OCR workflow that analyzes insurance forms for a client serving Indian hospitals.
mPragati Lab, IIT Delhi
Sept 2025 - Jan 2026
Developed and trained 3D U-Net pipelines for deep-learning-based skull implant reconstruction.
Coridors (Remote)
Aug 2024 - Sept 2025
Built a Snowflake-backed Streamlit ingestion platform with 30+ interactive screens and Snowpark workflows.
Delivered the company website and managed Vercel and Cloudflare deployments.
YouTube (@Param3021)
Jan 2023 - Present
Delivered 70+ live Python and ML sessions, reaching 175K+ learners.

A 1.2B parameter local coding assistant fine-tuned to run offline across CPUs and GPUs.
Fine-tuned LFM 2.5 with LoRA and scaled SFT context from 4K to 16K.
Integrated Ollama with Jupyter magic and a VS Code MCP server.

A production multi-agent system that researches repositories and writes personalized GitHub READMEs.
Orchestrated Detective, CTO, and Ghostwriter agents with LangGraph and real-time SSE.
Reached 100+ unique users in 16 hours and deployed on GCP Cloud Run.

A from-scratch LTC network for segmenting 32-dimensional motion time series.
Built custom LTC cells and reached 54.6% validation accuracy, outperforming the LSTM baseline by 9.6 points.

A scalable quiz platform with asynchronous jobs, rate limiting, and automated certificates.
Used Redis and Celery for distributed jobs, email workflows, and certificate generation.
Contact