Open to AI Data Ops & Evaluation roles

Manasjyoti Gogoi

AI Data Operations · LLM Evaluation · RLHF · AI Data Annotation

Manasjyoti Gogoi — AI Data Operations & LLM Evaluation
Professional Summary

Building the ground truth behind AI.

Computer Science graduate with hands-on experience in AI Data Operations, LLM evaluation, multimodal AI evaluation and annotation, Python, SQL, Power BI, and EDA. Experienced in rubric-based evaluation, pairwise model comparison, instruction-following assessment, rationale writing, ASR transcription, video-quality evaluation, and hierarchical robotic-action captioning across audio, video, speech, and visual tasks.

EF SET · C2 MasteryPearson Versant · C1
4.5/5
QA audit score on hierarchical robotic-action captioning
Multimodal
Audio, video, speech, and visual evaluation
RLHF
Pairwise, rubric, rationale & instruction following
Skills Matrix

A stack built for evaluation, data & delivery.

AI Evaluation & Data Ops

LLM EvaluationHuman Feedback EvaluationRubric-Based EvaluationData AnnotationPairwise ComparisonRationale EvaluationInstruction FollowingPrompt EvaluationAudio/Video EvaluationVideo SegmentationQA AuditsLogical Fallacy Identification

Programming & Analytics

PythonSQLC#Exploratory Data AnalysisData CleaningTranscription

Tools & Infrastructure

Power BIPower QueryDAXRoboflowComputer Vision LibrariesNumPyPandasMicrosoft ExcelLinuxGit
Experience

Where the work happens.

AI Data Annotator / AI Response Evaluator

Innodata

2026 — Present Remote
  • Conducted pairwise evaluation of live AI voice conversations — generating scenarios, scoring Model A vs. Model B on audio quality, conversational and emotional intelligence, and documenting timestamp-specific evidence with written rationales.
  • Performed long-form ASR transcription and speech annotation, labeling dialogue and non-speech events including breaths, coughs, hesitations, and inaudible segments.
  • Evaluated multimodal AI-agent interactions and video quality across camera quality, action density, cuts, lip synchronization, playback speed, and localized vs. global issues.
  • Performed hierarchical robotic-action captioning at L3 atomic-action, L2 object-action, and L1 whole-task levels — earning a 4.5/5 QA audit score for accuracy and guideline adherence.
Selected Projects

Shipping analytics and vision pipelines end-to-end.

Hyperion Insurance Operations & Sentiment Analytics

Interactive Power BI dashboard tracking coverage, claims, and premium performance across 10,000+ customer profiles.

600M+
Total Coverage
16.9M
Claim Expenses
5.9M
Premiums
  • Engineered ETL pipelines in Power Query to clean, transform, and map relational tables for 10,000+ customer profiles.
  • Built multi-dimensional DAX measures and cross-filtering visuals to segment policies and isolate demographic claim trends.
Power BISQLDAXPower Query

Warehouse Vision Pipeline & Dataset Curation

End-to-end computer vision data curation pipeline on Roboflow for multi-class industrial logistics datasets.

0.9%
Baseline mAP@50
3.5%
Precision
2.7%
Recall
  • Preprocessed, split, and labeled multi-class datasets using 2D localized bounding boxes.
  • Ran target-class distribution audits to identify environmental variation, class imbalance, and low-sample bottlenecks.
  • Optimized data-generation pipelines with class-balancing strategies to improve minority-class coverage.
RoboflowComputer VisionGitLinux
Education & Certifications

Formation & credentials.

Education

Kaziranga University
B.Tech, Computer Science and Engineering
Jorhat, Assam
Career Point Gurukul
Higher Secondary (Science)
Kota, Rajasthan
Bharatiya Vidya Bhavan
Secondary Education
Haldia, West Bengal

Certifications

AI Skills Passport
EY & Microsoft
AI + Sustainability Internship
1M1B
English · C1
Pearson Versant
English · C2 (Mastery)
EF SET
Python & SQL Certificates
CodeChef
Contact

Let's build the evaluation layer together.

Open to AI Data Ops, LLM evaluation, and annotation collaborations. Drop a note — I usually reply within 24 hours.