Anand K.

Key Skills

Software

Data Annotation Tech

Top Subject Matter

Call centre analytics and retail/petroleum customer interaction insights

Petroleum retail sales analytics via NL-to-SQL agent

Speech-to-text sequence generation for disfluency restoration

Top Data Types

Audio

Text

Document

Top Task Types

Transcription

Function Calling

Fine-tuning

Computer Programming/Coding

Evaluation/Rating

RLHF

Freelancer Overview

Call Insight AI — End-to-end call centre analytics pipeline (Whisper → intent classification → Gemini insights → SQLite . Brings 9+ years of professional experience across operations, contract review, compliance,B2B sales and structured analysis. Core strengths include HuggingFace and LangChain. Have experience of working on Data Annotation.tech. Education includes Bachelor of Science, IIT Madras (2025) and Bachelor of Engineering, BIT Mesra, Ranchi (2015). AI-training focus includes data types such as Audio, Computer Code, and Programming and labeling workflows including Transcription, Function Calling, and Fine-tuning.

Labeling Experience

Disfluency Restoration — Multimodal Seq2Seq (Wav2Vec2 + IndicBART)

AudioFine-tuning

Developed a multimodal seq2seq pipeline to restore disfluent transcripts from audio inputs. Extracted speech representations using Wav2Vec2 speech embeddings, fused them with text features, and fine-tuned IndicBART as the seq2seq generator to reproduce disfluent transcripts. Implemented a full end-to-end workflow including preprocessing, modality fusion, training, and inference. • Wav2Vec2 speech embedding extraction from audio • Modality fusion with text features • IndicBART fine-tuning for seq2seq generation • End-to-end preprocessing, training, and inference pipeline

2024 - Present

Sales Data Analyser — Natural language querying agent for petroleum retail CSVs

Function Calling

Built an NLP agent that enables plain-English querying over petroleum retail CSV data using LangChain and Gemini orchestration. The system translates user questions into SQL-backed answers and returns domain-aware responses grounded in outlet performance and volume trends. Focused on practical agent design to minimize the need for manual SQL by non-technical stakeholders. • LangChain + Gemini orchestration for NL → SQL • Domain-aware query generation for petroleum retail • Contextual response framing using business metrics • Agent-based interaction for analytics retrieval

2024 - Present

Call Insight AI — End-to-end call centre analytics pipeline (Whisper → intent classification → Gemini insights → SQLite → Streamlit)

AudioTranscription

Developed a call center analytics pipeline that transcribes customer audio and converts it into structured, downstream analytical signals. Used Whisper for transcription, then applied zero-shot classification to detect intents and Gemini API to generate insights, finally storing results in SQLite for reporting. Delivered the outputs through a Streamlit dashboard with modular components for filtering and drill-down. • Whisper transcription from call audio • Zero-shot intent classification from transcribed text • Gemini-based insight generation • SQLite + Streamlit reporting interface

2024 - Present

Education

Work History