Md Affan

Audio Analytics and Geospatial Intelligence projects leveraging Speech AI, NLP, LLMs, and advanced ML pipelines.

Data Scientist - MENRV.AI

(2024-08)

Audio Analytics and Geospatial Intelligence projects leveraging Speech AI, NLP, LLMs, and advanced ML pipelines.

• Developed Call Center, Meeting, and Lecture Analytics products leveraging Speech AI, NLP, LLMs, and audio processing pipelines.
• Built an end-to-end Audio Analytics pipeline covering Speech-to-Text (STT/ASR) with ~216× real-time processing (~1 min audio in ~0.28s) and ~8% WER, Text-to-Speech (TTS), speaker diarization (~12% DER), and voice cloning, leveraging Hugging Face Transformers and multiple APIs.
• Researched and prototyped an Agentic AI architecture integrating meeting and lecture analytics with live platforms (Google Meet, Microsoft Teams, Zoom) for automated transcription, speaker-level analysis, summarization, and action-item extraction.
• Automated structured logging, model evaluation, and PDF BI report generation for model explainability, performance monitoring, analytics insights, and business consumption.
• Developed a Salesman Routing & Optimization engine using clustering, geospatial analysis, and route optimization algorithms, reducing overall travel distance and ETA by >50% across multi-store routes.
• Contributed to Route Planning & Optimization pipelines for distance/ETA prediction, optimal path selection, and route recommendations, leveraging geospatial features, routing APIs, and ML-based ETA models.

Data Scientist - NEOPIC.AI

AI Image Platform development and integration.

• Developed and integrated GenAI image generation/editing pipelines, benchmarking 3+ foundation models/providers across multiple image-generation and editing use cases.
• Evaluated models using CLIP-based similarity, prompt adherence, visual consistency, latency, and cost, enabling use-case-specific model selection and routing.
• Built custom Computer Vision + GenAI workflows for Virtual Try-On (VTON), Apparel Recoloring (color extraction/segmentation), Infographic Generation, and domain-specific image editing using CAG.

Data Scientist (Internship) - MENRV.AI

(2023-09 - 2024-07)

Geospatial Intelligence and Audio Analytics projects.

• Developed a Driver-Target Assignment & Optimization pipeline using Haversine distance, geospatial clustering, and assignment algorithms, reducing total travel distance by ~65%.
• Contributed to an ML-based ETA prediction pipeline incorporating real-time features (weather, time window, day, geospatial attributes), achieving 2.7-min MAE and 0.91 R² for logistics operations.
• Automated PDF BI report generation for geospatial products, capturing model metrics, data validation, accuracy trends, and performance visualizations for every pipeline run.
• Developed Speaker Diarization, Speech Emotion Recognition (SER) using AVD scoring, and Sentiment Analysis modules for speech analytics using Hugging Face Transformer models.
• Collaborated with the backend team to build FastAPI-based REST APIs, optimize model inference, and develop production-ready ML services for multiple AI products.
• Evaluated and traced RAG/Agentic workflows using LangSmith, benchmarking retrieval relevance, answer correctness, faithfulness, latency, and token usage to optimize retrieval and LLM performance.

B.Tech - Mechanical Engineering - Jamia Millia Islamia (2020-12 - 2024-07)