Headline

Jr. Data Scientist @OFI Services | Final Year CS @ VIT Bhopal | Full Stack + AI/ML | Python · FastAPI · React.js · Next.js | AWS AI/ML Scholar · Microsoft Include Her

About

Building production AI systems | Full Stack + ML Engineering | Graduating Oct 2026 Currently: Machine Learning Engineer @ Manya Fintech — architecting a distributed 14-stage AI video intelligence pipeline processing 50+ match videos/week at 88% classification accuracy using Python, FastAPI, Redis, and Google Gemini. What I build: → AI-powered platforms (MediVox — 1,000+ users, 95% accuracy, sub-300ms latency) → Real-time distributed backends (FastAPI · Celery · Redis · PostgreSQL · WebSockets) → Full-stack web apps (React.js · Next.js · TypeScript · Tailwind CSS) → ML pipelines end-to-end (model training → MLflow tracking → Docker deployment) Stack: Python · React.js · Next.js · FastAPI · Node.js · PostgreSQL · Docker · AWS · Redis · MLflow Credentials: 🏅 AWS AI/ML Scholar 🏅 Microsoft Include Her Cohort 4.0 🏅 McKinsey Forward Program 🏅 Google for Startups — Prompt to Prototype 🏅 SIH 2023 Finalist Open to full-time SWE · ML Engineer · Full Stack roles (2026 fresher). 📩 abhimakulshrestha2004@gmail.com | portfolio-99lb.vercel.app

Location

Bengaluru

AI stack

aistudio

antigravity

claude code

claude

codex

cursor

Work experience

Junior Data Scientist at OFI Services, Jun 2026 - Present · 3 mos, Bengaluru

Machine Learning Engineer at TwynUp, Apr 2026 - Jun 2026 · 3 mos, Architecting an enterprise-scale AI sports analytics platform — distributed 14-stage video intelligence pipeline that outputs tactical reports, highlight clips, and CSV analytics using Google Gemini 2.5 Flash/Pro. Processing 50+ match videos/week at 88% event classification accuracy. Key work: • Scalable async backend: FastAPI · Celery · Redis · PostgreSQL — cut failure reruns by 60% and runtime by 35% • Automated FFmpeg highlight generation — eliminated 70% of manual editing effort • Real-time Server-Sent Events (SSE) progress tracking and interactive Next.js analytics dashboard • Trained custom OpenWakeWord wake-word models with MLflow-tracked evaluation pipelines • Implemented Intersection over Union (IoU) temporal deduplication — sub-2s timestamp sync accuracy across multi-camera feeds Stack: Python · FastAPI · Celery · Redis · PostgreSQL · Docker · Google Gemini · OpenCV · Next.js · MLflow

Software Development & Machine Learning Intern at TwynUp, Jan 2026 - Mar 2026 · 3 mos, Built real-time sports analytics dashboards (React.js · PostgreSQL · WebSocket) with <200ms latency across 3 concurrent feeds. Designed normalised database schemas and CI/CD pipelines via GitHub Actions — cutting deployment cycle by 25%. • Developed frontend and backend modules for sports analytics platforms • Contributed to automation workflows, database design, and early-stage event detection systems Stack: React.js · Python · PostgreSQL · WebSockets · GitHub Actions

Fullstack Development Intern at SignSetu, Oct 2025 - Nov 2025 · 2 mos, Hyderabad, Built accessible frontend and backend modules for a Deaf Education Technology platform serving 200+ learners — improving task completion by 30% and cutting the release cycle from 2 weeks to 4 days through Agile sprints and CI/CD automation. • Designed and implemented 5+ interactive React.js/Node.js modules improving usability and accessibility for Deaf and hard-of-hearing learners • Collaborated in Agile (Scrum) sprints to iteratively develop and deploy new features — reduced release cycle from 2 weeks to 4 days via CI/CD pipelines • Maintained scalable, well-documented codebases and contributed to version control workflows using Git and GitHub • Worked across full stack (frontend UI + backend APIs) to deliver responsive, inclusive user experiences Stack: React.js · Node.js · REST APIs · Git · GitHub · CI/CD · Agile (Scrum)

AI Intern at All India Council for Technical Education (AICTE), Jun 2023 - Jul 2023 · 2 mos, Remote, Built a Mental Health Fitness Tracker using Python and Linear Regression — applied GridSearchCV hyperparameter optimisation, improving model R² by 15% to achieve 82% stress-prediction accuracy on a 10-feature dataset. Stack: Python · Scikit-learn · Linear Regression · Feature Engineering

Education

Bachelor of Technology - BTech, Computer Science, Vellore Institute of Technology

Bansal Public School

Public links

github.com: https://github.com/https://github.com/abhimakulshrestha

portfolio-99lb.vercel.app: https://portfolio-99lb.vercel.app/