Dr. Manjeet Dahiya

VP of Engineering

Gurgaon, Haryana, India18 yrs 2 mos experience

AI EnabledAI ML Practitioner

Key Highlights

Developed foundational LLM reducing misroutes significantly.
Created personalized recommender systems for millions of users.
Led AI initiatives improving operational efficiencies at Ecom Express.

Stackforce AI infers this person is a Machine Learning and AI expert with extensive experience in logistics and automotive industries.

Contact

manjeetdahiya@gmail.com LinkedIn

Skills

Core Skills

Machine LearningArtificial IntelligenceData EngineeringData ScienceNlp

Other Skills

Agile MethodologiesAlgorithmsApache SparkArduinoBackend EngineeringBig DataC++Compiler ConstructionCompiler OptimizationCompilersCross-platform DevelopmentDeep LearningDesign PatternsDistributed SystemsEmbedded Systems

About

Dr. Manjeet Dahiya is the Vice President and Head of ML & Engineering at Ecom Express. Earlier, he headed ML and Data Engineering at CarDekho and ML at Airtel Digital. Manjeet obtained his PhD in computer science from IIT Delhi, and BTech in electrical engineering from IIT Kanpur. Over the years, Manjeet has led the development of AI systems for different industries. At Airtel Digital, he developed personalized recommender systems for Wynk Music catering to millions of users - Daily Mixes is a notable product. For CarDekho, he developed used cars pricing engine that in addition to predicting the right prices also influenced the P&L of the respective business unit. At Ecom Express, he leads the charter of increasing operational efficiencies and reducing costs through AI interventions. Toward this goal, he has developed an LLM from scratch for the address and location intelligence domain, which has been created into a stand-alone product called Bulls.ai. Manjeet has varied experience in different aspects of computer science, and his expertise are machine learning, generative AI, AI agents, deep learning, data engineering, backend engineering, programming languages, compilers, systems, and verification. He has published multiple research papers in academic and industrial settings.

Experience

Ecom express limited

VP & Head - Machine Learning & Engineering

Sep 2022 – Apr 2025 · 2 yrs 7 mos

Bulls.ai: Foundational LLM for Address and Location Intelligence. This resulted in reducing misroutes from 6.5% to 2%. 4.3% spelling corrections. 1.5% locality completion and 1% PIN correction. Detecting 0.5% invalid addresses. It is a decoder-only transformer model trained from scratch on address and delivery geo-coordinates. Trained 354M, 773M and 1.5B parameter models. This work received recognition from Aegis Graham Bell Awards & AWS AI Conclave.
Using product descriptions, identification of non-dangerous, liquid, and fragile goods to prevent losses and offloading in air cargo. This prevented losses & offloading (to just a few cases in a month) and TAT breaches. We achieved over 99% accuracy for these encoder-only transformer models.
PAN/Aadhaar document understanding: Improved FE onboarding experience by prefilling PAN/Aadhaar numbers and preventing incorrect entries. This allowed 82% of the onboarding without any human intervention.
FE/ESP payout optimization by simple ideas of deliveries for the same address, locality, and distance-based efforts. This led to savings of 30 paise per shipment cost.
Correct weight identification based on sorter images by detecting Box/Flyer. This led to an increase in revenue by Rs 2 cr per month.
Face-based attendance & helmet detection. Automated system to ensure compliance and safety on the ground.
Operations research: FLPs, Network planning, Milkrun design. This resulted in the automation of various operational planning. 86% automated DC location planning, 75% automated milkrun planning. TAT breach improvement by 9.7% through hub placement optimization

Machine LearningArtificial IntelligenceData EngineeringGenerative AIDeep LearningNatural Language Processing (NLP)

Cardekho

VP & Head - Machine Learning & Data Engineering

May 2021 – Sep 2022 · 1 yr 4 mos

Personalization: Listing and notifications personalization for optimizing conversion. This ranking system led to an improvement of 18% in test drive bookings.
Article: https://medium.com/@cardekhodatascience/the-impact-of-personalizing-the-experience-of-used-car-buyers-bd29db2f516e
Used car pricing: A neural network-based pricing model to output the price range of a used car based on the car's information and condition. This resulted in automating 70% of the used pricing without intervention.
Dynamic CTA selection: Automatic selection of CTA of lead flow buttons with the goal of CTR maximization. Modeled it as a bandit problem and applied the Epsilon-greedy reinforcement learning algorithm. The system led to an improvement of 15% in lead creation.
Data Engineering: Developed a central data lake platform for analytics and data science.

Machine LearningData EngineeringArtificial IntelligencePythonAlgorithms

Airtel digital

Head - Machine Learning & Data Sciences

Aug 2020 – May 2021 · 9 mos

Recommender systems catering to 66M user base in various form factors: personalized playlists curation, similar content, and artists and album recommendations. These systems led to a consumption of over 20% from recommendations alone.
Search personalization: The keyword “love” could mean “love you zindagi” or “love me like you do” for different users. Personalization increased search result consumption by 5% and up to 2x for long-tail, non-popular queries.
Content classification: Mood prediction from audio spectrograms. Catering to 1% of our search queries that looked for mood keywords.
Article: https://www.linkedin.com/pulse/ai-airtel-digital-manjeet-dahiya/

Machine LearningData ScienceArtificial IntelligenceNLP

Delhivery

Principal Data Scientist

Aug 2018 – Aug 2020 · 2 yrs · Gurgaon, Haryana, India

Multiple NLP and geospatial models on location intelligence and maps capabilities: address classification, address to geocode, address PIN correction, address normalization, address validity, learning locality/PIN polygons and learning roads.

NLPGeospatial IntelligenceData Engineering