TECH STACK

KEY RESULT

Reduced hallucinations · Personalized healthcare recommendations

METRICS

◈ Reduced hallucinated medical outputs by ~40% vs. zero-shot baseline

◈ Latency under 2.5s per recommendation

TECH STACK

KEY RESULT

Automated expense parsing · Real-time balance calculation

METRICS

◈ Receipt parsing accuracy >90% on test set

◈ Sub-200ms API response for balance calculation

TECH STACK

KEY RESULT

High accuracy · Robust to lighting variance

METRICS

◈ Validation accuracy: 97.3%

◈ False positive rate: <1.8%

TECH STACK

KEY RESULT

Highly personalized content discovery at scale

METRICS

◈ RMSE of 0.89 on test set

◈ Precision@10: 0.74

TECH STACK

KEY RESULT

Stable regression · Reduced prediction variance

METRICS

◈ R² of 0.81 on test set

◈ Ensemble reduced variance by ~18%

TECH STACK

KEY RESULT

7.9% WAPE · won 154/240 series · drift-triggered auto re-selection

METRICS

◈ Gradient-boosted model won 154/240 series at 7.9% WAPE

◈ 74.2% measured interval coverage vs. 80% nominal target

LLM Systems & RAG

Production-grade retrieval-augmented generation

Proven by

Preventive Care RAG Agent — 38% hallucination reduction on CDC corpus

ML Engineering

End-to-end model development & training

Proven by

BERT Product Predictor — ensemble of transformer + gradient boosting for regression

Computer Vision

Custom CNN architectures for visual tasks

Proven by

Fake Currency CNN — 97.3% accuracy on custom banknote dataset with 5-layer custom architecture

Recommender Systems

Collaborative filtering & latent factor models

Proven by

Netflix Recommender — SVD with 50 latent factors, Precision@10 of 0.74 on held-out set

AI Deployment & Infra

Containerised, production-ready AI systems

Proven by

All 6 projects deployed end-to-end with Docker + REST APIs, not just Jupyter notebooks

LLM Evaluation

Safety, accuracy & alignment assessment

Proven by

Handshake AI — evaluating LLM outputs for accuracy and safety in production systems

CS Foundations

Fell into Python at 3am debugging a loop that wouldn't end. Never stopped.

ML Fundamentals

First neural network. First overfit. First moment of understanding why regularisation exists.

Leadership + Workshops

GDSC Content Lead. Taught ML to 200+ students. Realised teaching clarifies your own thinking.

Production AI Systems

Built CNN for currency detection, Netflix recommender, first freelance ML contracts.

LLM Systems & CSUN

M.S. CS at CSUN. Built RAG pipelines and SplitWise AI.

Generative AI Analyst

Joined Handshake AI as a Generative AI Analyst — deep RAG workflows, production evaluation frameworks, LLM alignment research at CSUN.

Ready to Ship

M.S. complete May 2026. Full-time ML/AI engineer. Six production systems. Zero excuses.