Program Associate, Data Execution Role snapshot
• Role: Program Associate – Execution (Data Team)
•

Location:
Gurugram on-site, with field travel for some projects
•

Experience:
0–2 years (freshers with strong ops exposure welcome)
• Type: Full time
• Languages: Hindi and English required; a regional or foreign language is a plus

About the role
We build speech, text and video datasets in Indian and global languages that AI teams use to train and test voice models. As a Program Associate, you make sure the data for each client order gets collected, checked and delivered on time, at the quality the order specifies. This is an execution job, run at scale. You will not build the app, train models or design the dataset. You will source hundreds of speakers, listen to recordings, chase transcribers, reconcile QC sheets, pay freelancers and report numbers every day. Think of a 100-hour delivery target the way a plant manager thinks about a production line. What this job is not: a product, research, tech or pure coordination role. If you are mainly looking for those, this role will not be a good fit. What you will do Every project runs on the same contributor funnel: leads sourced → applied → test passed → hired → in group → working → work accepted. You own that funnel for your languages or regions, end to end.
• Recruit at scale. Run paid lead-gen and job-platform campaigns with the growth team, screen test recordings, and onboard speakers, transcribers and field workers. Report lead-to-working % weekly and fix the stage that is leaking.
• Check quality. Listen to recordings and run or supervise audio QC against the client's written spec (format, channels, noise, consent, no personal information). Log a reason for every rejection.
• Run transcription and annotation. Allocate work on the annotation tool, track output per person per day, audit samples, and block people who do not meet the bar.
• Know the guideline. Learn each client's recording, transcription and metadata guideline well enough to train a new contributor on it in one sitting, and to spot deviations in the data.
• Track and report. Update a daily project tracker: funnel, hours recorded / accepted / transcribed / delivered, cost per delivered hour, QC coverage. Send a short weekly status without being asked.
• Pay people correctly. Reconcile accepted work against payouts before every payment run.
• Protect data. Keep consent,



metadata and storage discipline on every single file.
• Deliver. Package data in the agreed folder structure with the counts sheet and a delivery note that discloses every assumption affecting price or quality.
• Escalate early. Flag a target that will be missed the week it becomes likely, not the week it is due. Projects you may work on You will be assigned to one or two of these workflows based on your languages, location and strengths. The core work above applies to all of them.
• Two-speaker conversational audio (Indian languages): pair speakers for natural phone conversations, write topic prompts, QC both speakers' audio. Office-based.
• Speech recognition (ASR) data: recruit against district and demographic quotas, manage transcription teams, track cost per hour. Office-based.
• Healthcare and domain field recordings: onboard frontline workers to record real conversations, with strict consent and privacy. Field travel.
• Tribal and low-resource languages: set up in a district, train community contributors on speech, translation and dictionary work. Relocation for ~3 months.
• Dialect data for institutional projects: turn a tender into a per-dialect plan and hit contractual milestones. Field coordination, north/central India.
• International accented English: launch new accents country by country and manage local leads and payouts. Remote, across time zones.
• Global-language conversational audio: launch languages on a calendar, QC two-channel audio, manage outside vendors. Remote, across time zones.
• Transcription quality: measure transcriber error rates (WER), find systematic errors and improve guidelines. Analytical.
• Face + voice video data: recruit and coach on-camera speakers, QC audio and video together. New process.
• Text-to-speech (TTS) data and evaluation: run studio sessions and manage hundreds of native-speaker raters. Studio + remote. How you will be measured (KRAs) You are measured on delivered hours, not recorded hours, and on what each delivered hour costs.
• Delivery: delivered hours per language or region each month, at or above plan
• Yield:



accepted vs recorded hours at or above baseline
• Funnel health: lead-to-working % tracked weekly, with a fix when it drops
• Cost: cost per delivered hour within budget; every payout reconciled
• Quality: client audits passed first time; zero batches returned
• Diversity and spec: gender, age, region and dialect targets met at every submission
• Compliance: consent and complete metadata on 100% of files; zero privacy incidents
• Timeliness: milestones met; risks flagged the week they appear
• Reporting: daily tracker current; weekly status sent without reminders Who we are looking for Must have
• Graduate in any discipline, with 0–2 years of experience.
• Some hands-on operations exposure where you were measured on output or quality: field programmes, campus or community mobilisation, BPO or annotation ops, logistics, QA, or a large college event you ran.
• Fluent Hindi and English, spoken and written.
• Comfort with numbers: funnels, yields, per-hour costs. You read the sheet before forming an opinion.
• Strong Google Sheets or Excel skills (filters, pivots, lookups).
• Stamina for repetitive, high-volume work (listening to 30–50 recordings a day, reconciling sheets, chasing people) over weeks, not days.
• Confidence managing large freelancer groups on WhatsApp or Telegram, including rejecting work and holding payouts.
• Attention to detail at the file level: a wrong sample rate, a missing channel, an unsigned consent.
• Willingness to travel or relocate for field projects, or to work odd hours for international ones.  Good to have
• A regional Indian language (essential for field projects), or a foreign language such as Spanish, French, Arabic or Mandarin.
• Reading knowledge of two or more Indian scripts.
• A good ear for pronunciation, accents and audio quality.
• Basic SQL or familiarity with metrics like WER/CER.
• Experience with paid social ads or job-platform recruitment. This role is not for you if
• You want to move into product, research or engineering soon.
• You are not comfortable with phone-heavy work, daily targets or field travel.
• You prefer coordinating meetings to owning a number. Hiring process
• Application and short screening call
• Take-home exercise that mirrors the real work
• Interview with the Program Manager
• Final conversation with the Data team lead

📌 Program Associate (Gurugram)
🏢 Josh Talks
📍 Gurugram