01
Normalized Evaluation
Metrics must be comparable. I build normalization layers to standardize disparate scores into unified scales for clear decisions.
Work
Selected projects that reflect my focus: clear metrics, reliable pipelines, and actionable AI insights.
PrepGenie
An LLM-powered interview coach utilizing Google Gemini 1.5 Flash and BERT. Delivers personalized mock interviews with real-time speech recognition and automated scoring across 6+ soft skills.
6+
Skills Scored
100%
Automated Feedback
Real-Time
Speech Processing
Actionable
Reports
Gemini
BERT
Gradio
Firebase
Speech Recognition
Technical Challenge
Making AI ask personalized questions. I crafted specific prompts for Gemini using resume summaries and job roles to generate relevant, non-technical questions focused on key soft skills.
EvalGenie
AI Content Quality Evaluator. Engineers a dynamic normalization layer using empirical metric ranges (BLEU, BERTScore) to standardize disparate NLP scores into a unified 0-1 scale.
7
Metrics Normalized
0-1
Unified Scale
Auditable
Quality Assessments
Provider-Neutral
Pipeline
Python
TensorFlow
NLTK
Gemini API
Groq API
Production Insight
Metrics aren't on the same scale. I created custom normalization ranges for each metric based on real-world performance data, making the final Hybrid Score meaningful and comparable.
Manifesto Explainer
Election Manifesto Analysis Platform. Uses NLP and Llama-3 to transform dense political documents into accessible insights, replacing keyword searches with contextual understanding.
10x
Faster Scanning
Contextual
Understanding
Summarized
Policy Themes
Deployed
Web App
Spring Boot
Python
NLTK
Llama-3
PostgreSQL
Technical Challenge
Manifestos are messy. I integrated Groq's Llama3-8b model to summarize what parties actually promise, turning static documents into interactive policy assistants.
GestureSynth
Real-time Hand Gesture to MIDI Conversion. A vision-based system using Python and MediaPipe that translates hand gestures into MIDI output with low latency.
30 FPS
Performance
<80ms
Latency
40%
Faster Learning
LLM-Adaptive
Tuning
Python
MediaPipe
OpenCV
MIDI
LLM
Technical Challenge
Calibrating gesture-to-music mapping in real-time. I built an LLM-powered adaptive tuning module that analyzes note stability metrics to automatically optimize parameters for improved playability.