VaradKhedkar KVarad777

"Until you make the unconscious conscious, it will direct your life and you will call it fate." — Carl Gustav Jung

✦ CORE ARCHITECTURE & BIO

▲ Academic & Engineering Matrix

B.Tech, Computer Engineering — VIT Pune (2025–2028)
Diploma, Information Technology — MSBTE Board (Govt. Polytechnic)
Focus: Machine Learning, Agentic Systems, AI Security & Indic NLP Research
Engineering Philosophy: Building deterministic, scalable, and self-healing AI systems.

◆ Industrial Leadership & Research Ops

IndicFoundation (Active Research): Script-agnostic phonological grounding framework across Indic languages.
Team Lead @ AlumiQA: Led a 4-engineer team building a hybrid commodity price forecasting engine.
AI Security Middleware: Architecting local PII redaction and local NER pipelines for secure cloud LLM inference.

❖ FEATURED SYSTEMS & RESEARCH PROJECTS

IndicFoundation — Script-Agnostic Multilingual Indic NLP

Unified multilingual research framework engineered to dismantle script-based barriers across Indian languages. Introduces a phonological grounding layer mapping written orthography to shared acoustic-phonemic representations, enabling cross-lingual knowledge transfer for low-resource languages.

Phonological Grounding Layer: Maps script variations across 22 scheduled Indian languages to a unified embedding space.
Joint Speech-Text Alignment: Integrated security modules defending against adversarial attacks and code-switched safety risks.

AlumiQA — Industrial Price Engine

Industrial-grade commodity price forecasting engine predicting aluminium market trends. Directed architecture design and team execution across 4 engineers, fusing GBDT time-series modeling with NLP-driven market text intelligence.

Coordinated data pipeline, modeling, and evaluation workstreams across a 4-member team.
Fused gradient-boosting time series with sentiment signals extracted from market financial reports.

LexSense — High-Performance NLP Search Indexer

High-throughput document retrieval and indexing engine written in native C++. Features custom NLP preprocessing algorithms, inverted index data structures, and rapid text scoring built for scale.

Native C++ implementation designed for minimal latency and memory overhead.
Custom tokenization pipeline and CMake build architecture.

Privacy Preserving Middleware

Security middleware inspecting API payloads for sensitive PII data prior to cloud LLM transmission. Enforces local entity recognition anonymization pipelines to ensure compliance and zero data leakage.

Local PII detection and automated anonymization before cloud egress.
Prompt injection defense and AI safety guardrails.

■ TECH ARSENAL & CAPABILITIES

Machine Learning, Deep Learning & NLP

Agentic AI Systems & LLM Orchestration

AI Security & Guardrails

Core Stack & Infrastructure

◆ DOMAIN MAP & OPERATIONS

[Active Multilingual Research]:
  - IndicFoundation: Script-Agnostic Phonological Grounding Layer
  - Cross-Lingual Knowledge Transfer Across 22 Scheduled Indian Languages
  - Joint Speech-Text Alignment & Code-Switched Adversarial Safety

[AI / ML Engineering]:
  - Predictive Time-Series Forecasting (GBDTs, LightGBM)
  - Feature Pipeline Architecture & Model Evaluation
  - Deep Learning Architectures (CNNs, LSTMs, Transformers)

[Agentic Systems]:
  - Multi-Agent Orchestration (LangChain, LangGraph, CrewAI)
  - Tool Calling & Memory Persistence
  - Retrieval-Augmented Generation (RAG) Systems

[AI Security & Guardrails]:
  - PII Detection & Local Anonymization Pipelines
  - Prompt Injection Mitigations & Model Egress Controls

"He who has a why to live can bear almost any how." — Friedrich Nietzsche

Navigation Menu