Skip to content

Open to Forward Deployed Engineer & AI Engineer roles

Hi, I'm Sahil Kumar.Building RAG and LLM systems, end to end.

Recent BS Data Science graduate (FAST NUCES, June 2026) focused on RAG, LLM fine-tuning and the data and cloud engineering needed to ship them.

  • RAG & LLMs
  • Python & SQL
  • PyTorch
  • FastAPI
  • Docker
View Projects Download Resume Contact Me
Portrait of Sahil Kumar

About

Recent graduate, building in public

I'm a recent Data Science graduate (BS, FAST NUCES, June 2026) who wants to work where AI meets real customer problems, which is why I'm targeting AI Engineer and Forward Deployed Engineer roles.

My focus is RAG and LLM applications: from simple retrieval to advanced, vectorless and graph-based RAG, with attention to evaluation, memory, security and guardrails, plus fine-tuning and knowledge distillation.

Since graduating I've been studying and building in focused sprints across machine learning and deep learning, data engineering and analytics, NLP, cloud, containers and system design. My project portfolio below is being built in public; each project shows its real status.

At FAST NUCES I worked as a Lab Demonstrator (DSA) and Teaching Assistant (Deep Learning, Parallel & Distributed Computing), explaining hard technical topics to other students.

3.84
CGPA / 4.00
3×
Gold Medalist
2.5 yrs
Teaching assistant experience

Projects

Project portfolio

I'm building these in public. Each card shows its real status; links appear only when the code or demo exists.

Data Science & MLPlanned

End-to-end ML pipeline: data collection to deployed API

Collect data, build an ETL pipeline into a database, run EDA, engineer features, train and evaluate a model, then serve it with separate model, backend and frontend layers.

  • ETL and storage in SQL
  • EDA and feature engineering
  • Metric tracking and improvement
  • API-served model with a frontend
  • Python
  • SQL
  • Pandas
  • scikit-learn
  • FastAPI
Data Science & MLPlanned

Business analytics dashboards (Power BI & Tableau)

Data cleaning, statistical analysis and visual storytelling on a public dataset, delivered as dashboards.

  • Cleaning and analysis
  • KPI dashboards
  • Written insights
  • SQL
  • Python
  • Power BI
  • Tableau
  • Excel
Data Science & MLCode on GitHub

Machine learning from scratch

Core ML algorithms implemented from first principles as part of a multi-day series.

  • Algorithms built without ML libraries
  • Python
  • NumPy
Code
Deep LearningPlanned

Time-series forecasting service

Forecasting with classical and neural models, evaluated properly and exposed through an API.

  • Baselines vs neural models
  • Backtesting
  • Deployed endpoint
  • Python
  • PyTorch
  • FastAPI
Deep LearningCode on GitHub

Image classification with FastAPI and Streamlit

A vision transformer model served behind a FastAPI backend with a Streamlit frontend.

  • Separated model, backend and frontend
  • PyTorch
  • FastAPI
  • Streamlit
Code
NLPCode on GitHub

RNN, gated RNN and LSTM text models

Gated RNNs from scratch and an LSTM sequence model for next-word prediction on IMDB reviews.

  • Gated RNN from scratch
  • Seq2seq next-word prediction
  • Python
  • PyTorch
Code
NLPPlanned

Deployed NLP application

A text classification or extraction pipeline covering core NLP concepts, packaged and deployed.

  • Tokenization to modeling
  • Containerized deployment
  • Python
  • Hugging Face
  • FastAPI
  • Docker
LLM Fine-tuningIn progress

Knowledge distillation into a small language model

Distilling a larger teacher model into a smaller student; early experiments are on GitHub.

  • Teacher-student training
  • Evaluation against the teacher
  • PyTorch
  • Hugging Face
  • PEFT
  • TRL
Code
LLM Fine-tuningPlanned

LoRA / QLoRA fine-tuning and a multimodal model

Parameter-efficient fine-tuning of an LLM, then a small vision-language model for document understanding.

  • SFT and DPO
  • Serving with vLLM
  • PyTorch
  • PEFT
  • TRL
  • vLLM
RAGCode on GitHub

FastAPI Docs Assistant: hybrid-search RAG with citations

Answers questions about the FastAPI docs from retrieved passages, cites its sources, and refuses when the docs don't cover the question. Not deployed yet; run it locally from the repo.

  • Hybrid retrieval (BM25 + dense, rank fusion): hit@5 1.00, MRR 0.89 on 30 hand-written questions, vs 0.90 / 0.77 for BM25 alone
  • Measured, then dropped, an off-the-shelf reranker that lowered MRR on this set
  • Calibrated a relevance guardrail on 38 questions; plus prompt-injection checks and rate limiting
  • 17 automated tests; evaluation is reproducible from the repo
  • Python
  • FastAPI
  • BM25
  • bge-small embeddings
  • Groq
  • Docker
Code
RAGCode on GitHub

Production-style RAG pipeline

A RAG pipeline using FastAPI, Inngest, Qdrant Cloud and Groq, designed to run on free-tier infrastructure.

  • Vector retrieval
  • Free-tier deployable design
  • FastAPI
  • Qdrant
  • Groq
  • Inngest
Code
RAGPlanned

Vectorless RAG and GraphRAG

Comparing vectorless and graph-based retrieval against vector RAG, with evaluation, memory and guardrails.

  • Eval comparison
  • Conversation memory
  • Security and guardrails
  • Python
  • LangGraph
  • Graph DB
  • RAGAS
Cloud & InfraPlanned

Containerized services on AWS and Kubernetes

Docker images deployed on AWS (EC2, S3, ECS/EKS) with Kubernetes manifests and RBAC, backed by SQL and NoSQL databases.

  • Infrastructure setup
  • RBAC
  • System design write-up
  • Docker
  • Kubernetes
  • AWS
  • PostgreSQL
  • MongoDB
  • Redis

More learning repos on GitHub

Open source

GitHub activity

A live snapshot of my public GitHub profile (taken Nov 2, 2026). My account started in September 2026, so the chart begins there.

  • 19

    Public repositories

  • 235

    Commits (last 12 months)

  • 35

    Longest daily streak (days)

  • 45

    Active days

265 contributions since I started

Best day: 33 on Oct 3

View as table (weekly totals)
Week ofContributions
Aug 30, 202627
Sep 6, 202626
Sep 13, 202651
Sep 20, 202637
Sep 27, 202660
Oct 4, 202629
Oct 11, 202622
Oct 18, 20265
Oct 25, 20260
Nov 1, 20268

Repositories by primary language

  • Python14
  • Jupyter Notebook4

Counts each public repo once by its main language (this website's own repo excluded).

Recently updated

Experience

Professional experience

  1. Jan 2024 – Jun 2026

    Lab Demonstrator (DSA) & Teaching Assistant (Deep Learning, PDC)

    FAST NUCES

    • Conducted Data Structures & Algorithms labs on complexity analysis, trees, graphs and dynamic programming with hands-on implementation.
    • Deep Learning TA: mentored students on CNNs, RNNs and Transformers in PyTorch, covering training, regularization and tuning; graded projects.
    • PDC TA: taught multithreading, MPI, OpenMP and CUDA, and reviewed assignments on distributed training and scaling compute-heavy workloads.
    • PyTorch
    • CUDA
    • MPI
    • OpenMP
    • DSA

Skills

Skills & expertise

Working knowledge

Used in coursework, teaching and public repositories.

Languages & Data

  • Python
  • SQL
  • C++

ML & Deep Learning

  • PyTorch
  • Hugging Face
  • CNNs / RNNs / Transformers

LLM & RAG

  • LangChain
  • LangGraph
  • LlamaIndex
  • Qdrant
  • FAISS
  • Chroma

APIs & Tools

  • FastAPI
  • Streamlit
  • Docker
  • Git & GitHub
  • PostgreSQL

Learning now

Actively studying and applying in current projects.

Analytics & BI

  • R
  • Power BI
  • Tableau
  • Statistics & EDA
  • Excel

LLM Engineering

  • LoRA / QLoRA / DPO
  • Knowledge distillation
  • Vectorless RAG
  • GraphRAG
  • RAG evals & guardrails

Cloud & DevOps

  • AWS (EC2, S3, ECS, EKS)
  • Azure
  • Kubernetes
  • Linux & Bash
  • CI/CD
  • Microsoft Foundry

Data Engineering & Systems

  • ETL pipelines
  • MySQL
  • MongoDB
  • Redis
  • Graph DBs
  • System design
  • Networking (TCP/IP, DNS, HTTP)

Now

Currently learning & building

  • Microsoft Foundry & Azure AI

    Starting hands-on work with Microsoft Foundry and related tooling.

  • Project sprint

    A two-month sprint turning what I've studied into deployed projects, added below as they ship.

  • Certifications

    Planning relevant cloud and AI certifications; none completed yet.

Education

Education & achievements

BS Data Science

FAST (NUCES) University · 2022 – June 2026

CGPA 3.84 / 4.00

Key courses

  • Machine Learning
  • Deep Learning
  • Agentic AI
  • MLOps
  • Computer Networks
  • Big Data Analytics

Achievements

  • Gold Medalist (3×)

Certifications

Pursuing relevant cloud and AI certifications; none completed yet.

Contact

Let's work together

Open to AI Engineer roles and collaborations. The fastest way to reach me is by email.

This opens your email app with the message pre-filled.