//Full-Stack DeveloperAI/ML Engineer
Building intelligent systems, beyond the prototype.
01 / Identity
I build AI systems that leave the lab.
I care about the gap between a promising demo and a system people can trust: grounded retrieval, honest evaluation, privacy by design, and infrastructure that survives real use.
01
LLMs, RAG pipelines, and computer vision — designed, trained, and evaluated for real workloads, not just benchmarks.
02
Three IEEE publications, and the habit of turning papers into shipped systems — ideas that survive contact with users.
03
FastAPI, Next.js, and the infrastructure around the models — the product surface that makes AI actually usable.
02 / Trajectory
Production AI · 2025
01 / 06
Lead AI & Full-Stack Developer · Regina, Canada
Led a privacy-first clinical AI platform and a multi-agent RAG knowledge assistant.
3–4×
Accuracy
<1.5s
Latency
70–80%
Fewer hallucinations
Worked with Python · TypeScript · FastAPI · Next.js · PostgreSQL · Docker
Publications · 2024 – 2025
02 / 06
Peer-reviewed · IEEE
Continual federated learning and telecom anomaly detection — three IEEE papers.
Better Continual Federated Learning using Generative Adversarial Networks
IEEE ICMLA · 2025
Continual Federated Learning for Dynamic Data Environments
IEEE PACRIM · 2024
Adaptive Thresholding Heuristic for KPI Anomaly Detection
IEEE COMSNETS · 2024
Telecom R&D · 2023
03 / 06
Research Intern · India
Designed a practical thresholding method for KPI anomaly detection on telecom data.
IEEE
COMSNETS '24
↓ noise
Fewer alerts
ML vs DL
Benchmarked
Worked with Python · PyTorch · Pandas · MLflow
Federated Learning · 2022
04 / 06
Research Intern · India
Implemented federated learning from scratch with continual-learning add-ons.
From scratch
FL impl.
EWC / LwF
Add-ons
Ablations
Documented
Worked with Python · PyTorch · TensorFlow · NumPy
On-device Vision · 2021
05 / 06
Research Intern · India
Prototyped monocular depth estimation tuned for mobile-device constraints.
Cert. of
Excellence
Monocular
Depth
On-device
Constraints
Worked with Python · PyTorch · TensorFlow · OpenCV
Continual Learning · 2020
06 / 06
Undergraduate Researcher · Singapore
Studied continual-learning methods and their stability–plasticity trade-offs.
EWC / LwF
Methods
2 yrs
Research
Singapore
Lab
Worked with Python · PyTorch · TensorFlow
03 / Deployments
Shipped products and research builds — scroll across the deck.
View all projects
Clinical AI · 2025
LiveAI clinical decision support for skin-lesion triage, clinician documentation, and referral workflows.
Built with Python · FastAPI · PyTorch · Next.js · PostgreSQL · MinIO
Marketplace · 2024
LiveOn-demand home-services marketplace with real-time booking and job management.
Built with Next.js · React Native · Node.js · Supabase
Productivity · 2024
LiveOffline-first productivity platform with a local AI assistant, file transfer, finance, and developer tools.
Built with TypeScript · React · Python · SQLite
Thesis · ICMLA 2025
ResearchGenerative-replay + EWC for long-horizon continual learning in federated systems. Accepted at IEEE ICMLA 2025.
Built with Python · PyTorch · MLflow
Telecom · COMSNETS 2024
ResearchAdaptive thresholding for telecom KPI anomaly detection — fewer noisy alerts, real faults preserved.
Built with Python · Pandas · MLflow
Agent Orchestration · 2025
PrototypeHierarchical-memory multi-agent system with vector search, streaming LLM responses, and CI/CD deployment.
Built with Next.js · FastAPI · Vector DB · LLMs
Real-time Systems · 2024
PrototypeModular real-time trading engine — candle ingestion, strategy execution, portfolio tracking, broker APIs.
Built with Python · WebSockets · Broker APIs · PostgreSQL
Let's build something
Regina --:--
© Akshat Sharma 2026
Download résumé